{"slug": "kimi", "name": "Kimi", "category": "ai", "category_name": "AI models and assistants", "description": "Moonshot AI's Kimi family of large language models, including open-weight releases like Kimi K3.", "recommenders": 87, "commenters": 161, "warners": 13, "score": 78.937, "rising": 14.35, "first": "2025-11-08", "latest": "2026-10-26", "why": "HN commenters use Kimi K3 as a daily driver for coding, planning and reviews, often comparing it to Opus-tier models, and value that it takes on security work that US models refuse.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 88, "neutral": 46, "compare": 27, "warn": 19}, "spark": [2, 1, 1, 1, 4, 1, 9, 16, 22, 16, 7, 8], "listed": true, "loved": ["Frontier-level quality", "Coding daily driver"], "caveat": "Cost and tokens", "short_quote": "To my surprise Kimi K2 is quite sharp, and often finds errors or omissions in the thinking and analyses of its colleagues.", "icon": true, "domain": "bearblog.dev", "github_repo": null, "official_url": "https://www.kimi.com", "pricing": "freemium", "aliases": ["Kimi", "Kimi K3", "Kimi K2.6", "Kimi K2.5", "Kimi K2", "Kimi K2.7", "Kimi 2.7", "Kimi 3", "Kimi 2.6", "Kimi K2.7 Code", "Kimi K3.1 Pro", "Kimi K.3", "Kimi-K3", "Kimi K2.7-Code", "Kimi-K2.6"], "verdict": [{"text": "HN commenters use Kimi K3 as a daily driver for coding, planning and reviews,", "cites": [{"id": 803877584, "item_id": 50242349, "url": "https://news.ycombinator.com/item?id=50242349", "by": "rglover", "date": "2026-10-26", "stance": "recommend", "quote": "Now it's my daily driver and I have zero complaints.", "excerpt": "\"Cheap\" is up for debate, but the models themselves are great. Kimi K3 (and now 3.1 Pro) was the first open model I played with and didn't think to switch back to a frontier model. Now it's my daily driver and I have zero complaints. The frontier labs are in a tough spot. It's hard to justify their cost when you can get something that's easily as good for cheaper (and anecdotally, more reliable/accessible). I do HITL", "context": "does human-in-the-loop work", "use_case": "daily driver for coding", "switched_from": "", "story": "IBM CEO says open weight models are now good and cheap on-prem"}, {"id": 789452800, "item_id": 49340800, "url": "https://news.ycombinator.com/item?id=49340800", "by": "Scene_Cast2", "date": "2026-08-17", "stance": "recommend", "quote": "Fwiw I love K3 and use it as a daily driver.", "excerpt": "Oh hey, that's cheaper than Kimi K3! Amusing to see a SOTA OpenAI model be cheaper than a Chinese open weight model. Fwiw I love K3 and use it as a daily driver. I haven't tried Sol, as I dislike OpenAI.", "context": "", "use_case": "daily driver", "switched_from": "", "story": "GPT-5.6 Sol Pricing Cut by 50%"}, {"id": 790522192, "item_id": 49407637, "url": "https://news.ycombinator.com/item?id=49407637", "by": "pimeys", "date": "2026-08-23", "stance": "recommend", "quote": "Except a month ago I started testing Kimi K3 and omp and I never went back.", "excerpt": "I've been using all the SOTA models a lot at work, like serious amount of tokens. It's been really rare that I stick with one model and harness for too long... Except a month ago I started testing Kimi K3 and omp and I never went back. Something with this combo works really well for Rust dev. The model doesn't really annoy me at all and I have not switched to Opus or SOL. And the monthly token bill is much lower...", "context": "heavy work user of SOTA models", "use_case": "Rust development", "switched_from": "", "story": "A week of using Codex more than Claude"}, {"id": 783994737, "item_id": 48999671, "url": "https://news.ycombinator.com/item?id=48999671", "by": "apatheticonion", "date": "2026-07-21", "stance": "recommend", "quote": "now Kimi K3 offers a great planning assistant for more advanced coding tasks", "excerpt": "I love the Chinese models. I use DeepSeek exclusively and now Kimi K3 offers a great planning assistant for more advanced coding tasks. DeepSeek v4 Flash is extremely fast and is able to handle pretty much anything I've thrown at it (I use mostly Rust, PSQL, Angular and Terraform). I self host Bifrost as my LLM gateway, though I wish LLM vendors would do monthly/daily automatic billing (like VPS providers do) rather ", "context": "", "use_case": "planning for coding", "switched_from": "", "story": "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"}, {"id": 792748850, "item_id": 49546803, "url": "https://news.ycombinator.com/item?id=49546803", "by": "nxdmum", "date": "2026-09-03", "stance": "recommend", "quote": "Combined with Kimi K3 for adversarial plan reviews - you get robust plans .", "excerpt": "so far using Gemini from 2.5 Pro to date (3.7 flash) - the way google trains the model it seems - is to identify top 3 to 4 things to fix first. as a result gemini is not that good in being thorough - but it's a needle mover . Opus5 Opus4.8 and Fable always point out things that Gemini missed. but Gemini was a needle mover - identifying the most important things to fix. I always enjoy interacting with Gemini . When i", "context": "", "use_case": "planning and plan review", "switched_from": "", "story": "Gemini 3.8 Flash and 3.8 Flash Cyber"}]}, {"text": " often comparing it to Opus-tier models,", "cites": [{"id": 784992032, "item_id": 49062002, "url": "https://news.ycombinator.com/item?id=49062002", "by": "Stagnant", "date": "2026-07-26", "stance": "recommend", "quote": "it is the first of the chinese models that actually feels comparable to Opus-tier models", "excerpt": "Yeah can't recommend their $19 plan, only took me a day and a half to hit the weekly usage cap. The $39 plan has 5x the limits so I recommend getting that instead. Having now tested it for a few days, it is the first of the chinese models that actually feels comparable to Opus-tier models", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Kimi K3 is not cheap"}, {"id": 785468576, "item_id": 49091786, "url": "https://news.ycombinator.com/item?id=49091786", "by": "rglover", "date": "2026-07-28", "stance": "recommend", "quote": "Just tried K3 out for the first time today and it's a legitimate threat.", "excerpt": "Just tried K3 out for the first time today and it's a legitimate threat. Temporarily (maybe permanently) using it as my daily driver but it's wild how comparable it is to Opus 4.7/4.8 (what's been my go to for a bit now\u2014wrote a quick post on what I found today [1]). [1] https://graybearding.bearblog.dev/kimi-k3-is-insane/", "context": "", "use_case": "daily driver coding", "switched_from": "Opus", "story": "Kimi K3 Architecture Overview and Notes"}, {"id": 797043536, "item_id": 49815221, "url": "https://news.ycombinator.com/item?id=49815221", "by": "abalashov", "date": "2026-09-23", "stance": "recommend", "quote": "I would generally consider the SWE capabilities of Kimi to be on par with Claude", "excerpt": "Data points from my own usage, for whatever they're worth[1]. I'm just an individual contributor and hardly represent enterprise or large orgs: I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! 1) For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit i", "context": "individual contributor", "use_case": "everyday coding", "switched_from": "", "story": "The current balance of power in open models"}]}, {"text": " and value that it takes on security work that US models refuse.", "cites": [{"id": 784869632, "item_id": 49054352, "url": "https://news.ycombinator.com/item?id=49054352", "by": "alexsmirnov", "date": "2026-07-25", "stance": "recommend", "quote": "No rejections, they did find the same problems on my project that I used for testing.", "excerpt": "I did run both kimi-k3 and glm-5.2 with Capital One vulnhunt [1]. No rejections, they did find the same problems on my project that I used for testing. gpt-5.6, gemini pro, and opus all rejected to follow. gpt even declined to edit skill files. [1] https://www.capitalone.com/tech/open-source/announcing-vulnhunter/", "context": "", "use_case": "vulnerability hunting", "switched_from": "", "story": "UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities"}, {"id": 792973745, "item_id": 49560859, "url": "https://news.ycombinator.com/item?id=49560859", "by": "DaSHacka", "date": "2026-09-04", "stance": "recommend", "quote": "I just switch the model to Kimi K3 to finish up making the PoC", "excerpt": "Surprisingly, I've had really good luck with reverse engineering on frontier models (without being part of CVP or similar). It's more the exploit development/PoC that it locks up on, which, as I use `pi`, I just switch the model to Kimi K3 to finish up making the PoC. Ironically, due to the stringent guardrails on American models that exist to avoid giving adversaries a leg-up in cybersecurity, I end up feeding dozen", "context": "", "use_case": "exploit PoC development", "switched_from": "", "story": "GPT-6 Astra"}, {"id": 788722627, "item_id": 49295164, "url": "https://news.ycombinator.com/item?id=49295164", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "So, I switched to Kimi K3 and DeepSeek for API testing just because it's so much cheaper.", "excerpt": "I've been building a harness for security work, and had to switch to GPT 5.5 when even Opus started refusing security work. Then 5.6 Sol arrived, and it refuses security work, too. So, I switched to Kimi K3 and DeepSeek for API testing just because it's so much cheaper. But, if GLM is better, I'm here for it, as I think GLM is also cheaper than K3.", "context": "building a harness for security work", "use_case": "API testing", "switched_from": "GPT 5.6 Sol", "story": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities"}, {"id": 783809504, "item_id": 48988094, "url": "https://news.ycombinator.com/item?id=48988094", "by": "mempko", "date": "2026-07-20", "stance": "recommend", "quote": "Switched to K3 and continued my work. And it worked really well for what I was doing.", "excerpt": "I can second this. I was using Fable for some research and fable said it wasn't going to do it and downgraded to opus. Switched to K3 and continued my work. And it worked really well for what I was doing. You can easily switch claude code to use K3, it took like 30 seconds.", "context": "", "use_case": "research, as Claude Code backend", "switched_from": "Fable", "story": "Kimi Work"}]}, {"text": " The most common complaint is high token usage and cost,", "cites": [{"id": 803830656, "item_id": 50239416, "url": "https://news.ycombinator.com/item?id=50239416", "by": "ipsum2", "date": "2026-10-26", "stance": "warn", "quote": "Kimi is by far the most expensive in terms of tokens and $/work done.", "excerpt": "> much cheaper to access I'm calling BS. I use Kimi, chatGPT (Astra/Sol), and Claude (Opus 5.5), and Kimi is by far the most expensive in terms of tokens and $/work done. Using OpenCode harness.", "context": "", "use_case": "coding work", "switched_from": "", "story": "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"}, {"id": 788624929, "item_id": 49289058, "url": "https://news.ycombinator.com/item?id=49289058", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "Kimi K3 is smarter, though.", "excerpt": "DeepSeek is more fun, because it's cheap-as-free, Good Enough, quite fast. I use it for all API stuff, automated runs, testing of the security auditing harness and benchmarks I'm working on, etc. Kimi K3 is smarter, though. At least smarter than DeepSeek V4 Flash 0731. I haven't tried the new Pro version, but will this weekend when I'm working on my personal projects But, K3 has been what I've been using for the actu", "context": "building a security auditing harness and benchmarks", "use_case": "coding a security auditing harness", "switched_from": "Claude", "story": "Qwen/Qwen3.8-2.4T-A95B"}, {"id": 788546979, "item_id": 49284186, "url": "https://news.ycombinator.com/item?id=49284186", "by": "gboss", "date": "2026-08-13", "stance": "warn", "quote": "Kimi 3 was super expensive as spec model", "excerpt": "For a personal project I\u2019ve been piloting Spec driven development (SDD), (it\u2019s contagious!), using Cursor and EARS statements. The strategy has been to use the frontier model to write the spec and a lessor model to write the tests and code and the frontier model to write critiquing prompts until it has nothing left to say. For my particular project there are two programs (or in human speak phases), where each program", "context": "", "use_case": "spec writing", "switched_from": "", "story": "Grok 4.6"}, {"id": 803445840, "item_id": 50215365, "url": "https://news.ycombinator.com/item?id=50215365", "by": "3princip", "date": "2026-10-23", "stance": "warn", "quote": "Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "excerpt": "Same plan, just asked it to review a repo that Fable has been helping out with (max plan, haven't run out of the hourly or weekly limit once yet, and I use the same sub on multiple projects at once). Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "context": "", "use_case": "reviewing a repo", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}]}, {"text": " and some say quality has declined since version 2.7.", "cites": [{"id": 798302643, "item_id": 49893915, "url": "https://news.ycombinator.com/item?id=49893915", "by": "dools", "date": "2026-09-29", "stance": "warn", "quote": "since 2.7 Kimi has gone way downhill", "excerpt": "I go for weeks without topping up my Deepseek balance, and have to top up my OpenAI balance at least every week. GPT 6 Luna closed the gap significantly for sure (it seems to be about twice as expensive as DS v4.1 Flash), but Deepseek v4.1 Flash is still the best value model and capable enough for almost everything I need to do. Sometimes if it's babbling or can't nail down a solution I switch to Sol for one prompt, ", "context": "formerly a heavy Kimi k2.5/2.6 user", "use_case": "", "switched_from": "", "story": "Sonnet 5.5"}, {"id": 787474369, "item_id": 49217148, "url": "https://news.ycombinator.com/item?id=49217148", "by": "dools", "date": "2026-08-07", "stance": "warn", "quote": "I was using Kimi a lot but it just nose dived.", "excerpt": "I have been using deepseek v4 pro almost exclusively. I was using Kimi a lot but it just nose dived. The decline started with the release of 2.7 and accelerated with the release of 3. When I need vision capabilities I use GPT 5.3 codex and if deepseek can\u2019t figure something out after a few goes I switch to GTP 5.5 or 5.6 (I\u2019ve been giving Terra first bite recently and it does pretty well, and have used Sol a couple o", "context": "", "use_case": "", "switched_from": "", "story": "DeepSeek V4 Flash 0731"}]}], "praise": [{"label": "Frontier-level quality", "summary": "Several commenters find it comparable to Opus or Claude and see little gap with frontier models.", "quotes": [{"id": 784992032, "item_id": 49062002, "url": "https://news.ycombinator.com/item?id=49062002", "by": "Stagnant", "date": "2026-07-26", "stance": "recommend", "quote": "it is the first of the chinese models that actually feels comparable to Opus-tier models", "excerpt": "Yeah can't recommend their $19 plan, only took me a day and a half to hit the weekly usage cap. The $39 plan has 5x the limits so I recommend getting that instead. Having now tested it for a few days, it is the first of the chinese models that actually feels comparable to Opus-tier models", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Kimi K3 is not cheap"}, {"id": 785468576, "item_id": 49091786, "url": "https://news.ycombinator.com/item?id=49091786", "by": "rglover", "date": "2026-07-28", "stance": "recommend", "quote": "Just tried K3 out for the first time today and it's a legitimate threat.", "excerpt": "Just tried K3 out for the first time today and it's a legitimate threat. Temporarily (maybe permanently) using it as my daily driver but it's wild how comparable it is to Opus 4.7/4.8 (what's been my go to for a bit now\u2014wrote a quick post on what I found today [1]). [1] https://graybearding.bearblog.dev/kimi-k3-is-insane/", "context": "", "use_case": "daily driver coding", "switched_from": "Opus", "story": "Kimi K3 Architecture Overview and Notes"}, {"id": 797043536, "item_id": 49815221, "url": "https://news.ycombinator.com/item?id=49815221", "by": "abalashov", "date": "2026-09-23", "stance": "recommend", "quote": "I would generally consider the SWE capabilities of Kimi to be on par with Claude", "excerpt": "Data points from my own usage, for whatever they're worth[1]. I'm just an individual contributor and hardly represent enterprise or large orgs: I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! 1) For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit i", "context": "individual contributor", "use_case": "everyday coding", "switched_from": "", "story": "The current balance of power in open models"}]}, {"label": "Coding daily driver", "summary": "Commenters have replaced Anthropic or other frontier models with it for everyday coding.", "quotes": [{"id": 803877584, "item_id": 50242349, "url": "https://news.ycombinator.com/item?id=50242349", "by": "rglover", "date": "2026-10-26", "stance": "recommend", "quote": "Now it's my daily driver and I have zero complaints.", "excerpt": "\"Cheap\" is up for debate, but the models themselves are great. Kimi K3 (and now 3.1 Pro) was the first open model I played with and didn't think to switch back to a frontier model. Now it's my daily driver and I have zero complaints. The frontier labs are in a tough spot. It's hard to justify their cost when you can get something that's easily as good for cheaper (and anecdotally, more reliable/accessible). I do HITL", "context": "does human-in-the-loop work", "use_case": "daily driver for coding", "switched_from": "", "story": "IBM CEO says open weight models are now good and cheap on-prem"}, {"id": 790522192, "item_id": 49407637, "url": "https://news.ycombinator.com/item?id=49407637", "by": "pimeys", "date": "2026-08-23", "stance": "recommend", "quote": "Except a month ago I started testing Kimi K3 and omp and I never went back.", "excerpt": "I've been using all the SOTA models a lot at work, like serious amount of tokens. It's been really rare that I stick with one model and harness for too long... Except a month ago I started testing Kimi K3 and omp and I never went back. Something with this combo works really well for Rust dev. The model doesn't really annoy me at all and I have not switched to Opus or SOL. And the monthly token bill is much lower...", "context": "heavy work user of SOTA models", "use_case": "Rust development", "switched_from": "", "story": "A week of using Codex more than Claude"}, {"id": 790988192, "item_id": 49436762, "url": "https://news.ycombinator.com/item?id=49436762", "by": "jazzpush2", "date": "2026-08-25", "stance": "recommend", "quote": "I have completed stopped using Anthropic since Kimi 3 released, and only ever K3 and Codex nowadays", "excerpt": "I have completed stopped using Anthropic since Kimi 3 released, and only ever K3 and Codex nowadays. I pay for K3, and get Codex and Claude free from work. Almost everyone I know prefers Codex to Claude.", "context": "", "use_case": "coding", "switched_from": "Claude", "story": "Anthropic Sees over $30T in Potential Revenue"}]}, {"label": "Fewer refusals", "summary": "It completes security and exploit work that GPT, Gemini and Opus refused.", "quotes": [{"id": 784869632, "item_id": 49054352, "url": "https://news.ycombinator.com/item?id=49054352", "by": "alexsmirnov", "date": "2026-07-25", "stance": "recommend", "quote": "No rejections, they did find the same problems on my project that I used for testing.", "excerpt": "I did run both kimi-k3 and glm-5.2 with Capital One vulnhunt [1]. No rejections, they did find the same problems on my project that I used for testing. gpt-5.6, gemini pro, and opus all rejected to follow. gpt even declined to edit skill files. [1] https://www.capitalone.com/tech/open-source/announcing-vulnhunter/", "context": "", "use_case": "vulnerability hunting", "switched_from": "", "story": "UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities"}, {"id": 792973745, "item_id": 49560859, "url": "https://news.ycombinator.com/item?id=49560859", "by": "DaSHacka", "date": "2026-09-04", "stance": "recommend", "quote": "I just switch the model to Kimi K3 to finish up making the PoC", "excerpt": "Surprisingly, I've had really good luck with reverse engineering on frontier models (without being part of CVP or similar). It's more the exploit development/PoC that it locks up on, which, as I use `pi`, I just switch the model to Kimi K3 to finish up making the PoC. Ironically, due to the stringent guardrails on American models that exist to avoid giving adversaries a leg-up in cybersecurity, I end up feeding dozen", "context": "", "use_case": "exploit PoC development", "switched_from": "", "story": "GPT-6 Astra"}, {"id": 783809504, "item_id": 48988094, "url": "https://news.ycombinator.com/item?id=48988094", "by": "mempko", "date": "2026-07-20", "stance": "recommend", "quote": "Switched to K3 and continued my work. And it worked really well for what I was doing.", "excerpt": "I can second this. I was using Fable for some research and fable said it wasn't going to do it and downgraded to opus. Switched to K3 and continued my work. And it worked really well for what I was doing. You can easily switch claude code to use K3, it took like 30 seconds.", "context": "", "use_case": "research, as Claude Code backend", "switched_from": "Fable", "story": "Kimi Work"}]}, {"label": "Planning and review", "summary": "Used for adversarial plan reviews, research, critique and finding bugs in old code.", "quotes": [{"id": 792748850, "item_id": 49546803, "url": "https://news.ycombinator.com/item?id=49546803", "by": "nxdmum", "date": "2026-09-03", "stance": "recommend", "quote": "Combined with Kimi K3 for adversarial plan reviews - you get robust plans .", "excerpt": "so far using Gemini from 2.5 Pro to date (3.7 flash) - the way google trains the model it seems - is to identify top 3 to 4 things to fix first. as a result gemini is not that good in being thorough - but it's a needle mover . Opus5 Opus4.8 and Fable always point out things that Gemini missed. but Gemini was a needle mover - identifying the most important things to fix. I always enjoy interacting with Gemini . When i", "context": "", "use_case": "planning and plan review", "switched_from": "", "story": "Gemini 3.8 Flash and 3.8 Flash Cyber"}, {"id": 783994737, "item_id": 48999671, "url": "https://news.ycombinator.com/item?id=48999671", "by": "apatheticonion", "date": "2026-07-21", "stance": "recommend", "quote": "now Kimi K3 offers a great planning assistant for more advanced coding tasks", "excerpt": "I love the Chinese models. I use DeepSeek exclusively and now Kimi K3 offers a great planning assistant for more advanced coding tasks. DeepSeek v4 Flash is extremely fast and is able to handle pretty much anything I've thrown at it (I use mostly Rust, PSQL, Angular and Terraform). I self host Bifrost as my LLM gateway, though I wish LLM vendors would do monthly/daily automatic billing (like VPS providers do) rather ", "context": "", "use_case": "planning for coding", "switched_from": "", "story": "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"}, {"id": 800181554, "item_id": 50011347, "url": "https://news.ycombinator.com/item?id=50011347", "by": "lmf4lol", "date": "2026-10-08", "stance": "recommend", "quote": "GLM-5.3 for coding or Kimi K.3 for research and critique.", "excerpt": "Oh man. v4.1-flash has been an sbolute game changer for us. We run all our Personal Assistants now on flash (thinking high) by default and it works incredibly well. There is really no need for basic agentic tasks that might require Kimi K.3 or GLM-5.3 levels. Once its gets juicier, we let flash launch specialized subagents with specific models. GLM-5.3 for coding or Kimi K.3 for research and critique. But as a main d", "context": "", "use_case": "research and critique", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}]}, {"label": "Open weights", "summary": "Commenters value that it is an open model they can run through multiple providers.", "quotes": [{"id": 802264432, "item_id": 50141527, "url": "https://news.ycombinator.com/item?id=50141527", "by": "verdverm", "date": "2026-10-18", "stance": "recommend", "quote": "models like Kimi K3 are right there, open and free to use, as an alternative to the closed models", "excerpt": "> I think it's still too early ... The data shows a clear trend, China is also doing the self improvement. From what we can tell, they are making the greater efficiency gains out of necessity (lack of easy access to the best hardware). Distillation is less important and models like Kimi K3 are right there, open and free to use, as an alternative to the closed models. I expect open weights to bypass closed models and ", "context": "", "use_case": "open-weights alternative to closed models", "switched_from": "", "story": "How the AI boom might end: 5 scenarios"}, {"id": 790570097, "item_id": 49410631, "url": "https://news.ycombinator.com/item?id=49410631", "by": "microtonal", "date": "2026-08-23", "stance": "recommend", "quote": "I have been using GLM and Kimi models the last few months mixed with the latest Anthropic models", "excerpt": "peace of mind versus using some Chinese cloud These are open weight models (GLM-5.3 soon too). You can run them on the Together AIs or Firework AIs of this world. Use OpenRouter or HF Inference Providers in between and you can effortlessly switch between models and providers. I have been using GLM and Kimi models the last few months mixed with the latest Anthropic models and for my daily work there is barely a differ", "context": "", "use_case": "daily work", "switched_from": "", "story": "GLM-5.3 (open-weight) beat Anthropic/OpenAI models \u2013 for 1/5 the cost"}, {"id": 789188624, "item_id": 49324289, "url": "https://news.ycombinator.com/item?id=49324289", "by": "alexjplant", "date": "2026-08-16", "stance": "recommend", "quote": "Check out Kimi, DeepSeek, and GLM series models in particular", "excerpt": "If you don't have regulatory hurdles to jump over then there are a whole raft of open-weight models that perform basically as well as Anthropic's offerings (and, in many cases, better because they're pulling fewer capacity-stretching shenanigans). Check out Kimi, DeepSeek, and GLM series models in particular with something like OpenCode or Pi as a coding harness. If you need a Claude-alike web UI then spin up Open We", "context": "", "use_case": "open-weight alternative to Claude", "switched_from": "", "story": "Claude Seems Down"}]}, {"label": "Writing style", "summary": "One commenter praises its writing style enough to use it to clean up verbose comments.", "quotes": [{"id": 803469888, "item_id": 50216868, "url": "https://news.ycombinator.com/item?id=50216868", "by": "simondotau", "date": "2026-10-23", "stance": "recommend", "quote": "Kimi K3 has been my go-to model for its writing style", "excerpt": "Kimi K3 has been my go-to model for its writing style and (prior to Muse Wonder) my go-to for a third opinion during multi-round/multi-agent analysis tasks. Its writing style is IMHO so good that, for code which humans need to review, I regularly got it to clean up the painfully verbose comments left by Opus. I'm excited to see if its writing style and unique takes have been retained in this new release.", "context": "", "use_case": "writing and multi-agent analysis", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}]}], "caveats": [{"label": "Cost and tokens", "summary": "Some find it expensive per unit of work, and one hit a plan's 5-hour budget in minutes.", "quotes": [{"id": 803830656, "item_id": 50239416, "url": "https://news.ycombinator.com/item?id=50239416", "by": "ipsum2", "date": "2026-10-26", "stance": "warn", "quote": "Kimi is by far the most expensive in terms of tokens and $/work done.", "excerpt": "> much cheaper to access I'm calling BS. I use Kimi, chatGPT (Astra/Sol), and Claude (Opus 5.5), and Kimi is by far the most expensive in terms of tokens and $/work done. Using OpenCode harness.", "context": "", "use_case": "coding work", "switched_from": "", "story": "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"}, {"id": 788624929, "item_id": 49289058, "url": "https://news.ycombinator.com/item?id=49289058", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "Kimi K3 is smarter, though.", "excerpt": "DeepSeek is more fun, because it's cheap-as-free, Good Enough, quite fast. I use it for all API stuff, automated runs, testing of the security auditing harness and benchmarks I'm working on, etc. Kimi K3 is smarter, though. At least smarter than DeepSeek V4 Flash 0731. I haven't tried the new Pro version, but will this weekend when I'm working on my personal projects But, K3 has been what I've been using for the actu", "context": "building a security auditing harness and benchmarks", "use_case": "coding a security auditing harness", "switched_from": "Claude", "story": "Qwen/Qwen3.8-2.4T-A95B"}, {"id": 803445840, "item_id": 50215365, "url": "https://news.ycombinator.com/item?id=50215365", "by": "3princip", "date": "2026-10-23", "stance": "warn", "quote": "Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "excerpt": "Same plan, just asked it to review a repo that Fable has been helping out with (max plan, haven't run out of the hourly or weekly limit once yet, and I use the same sub on multiple projects at once). Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "context": "", "use_case": "reviewing a repo", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}]}, {"label": "Quality decline", "summary": "Former heavy users say it got worse starting with version 2.7.", "quotes": [{"id": 798302643, "item_id": 49893915, "url": "https://news.ycombinator.com/item?id=49893915", "by": "dools", "date": "2026-09-29", "stance": "warn", "quote": "since 2.7 Kimi has gone way downhill", "excerpt": "I go for weeks without topping up my Deepseek balance, and have to top up my OpenAI balance at least every week. GPT 6 Luna closed the gap significantly for sure (it seems to be about twice as expensive as DS v4.1 Flash), but Deepseek v4.1 Flash is still the best value model and capable enough for almost everything I need to do. Sometimes if it's babbling or can't nail down a solution I switch to Sol for one prompt, ", "context": "formerly a heavy Kimi k2.5/2.6 user", "use_case": "", "switched_from": "", "story": "Sonnet 5.5"}, {"id": 787474369, "item_id": 49217148, "url": "https://news.ycombinator.com/item?id=49217148", "by": "dools", "date": "2026-08-07", "stance": "warn", "quote": "I was using Kimi a lot but it just nose dived.", "excerpt": "I have been using deepseek v4 pro almost exclusively. I was using Kimi a lot but it just nose dived. The decline started with the release of 2.7 and accelerated with the release of 3. When I need vision capabilities I use GPT 5.3 codex and if deepseek can\u2019t figure something out after a few goes I switch to GTP 5.5 or 5.6 (I\u2019ve been giving Terra first bite recently and it does pretty well, and have used Sol a couple o", "context": "", "use_case": "", "switched_from": "", "story": "DeepSeek V4 Flash 0731"}]}, {"label": "Inconsistent output", "summary": "One commenter finds output quality varies more than with Anthropic or OpenAI models.", "quotes": [{"id": 787223538, "item_id": 49201471, "url": "https://news.ycombinator.com/item?id=49201471", "by": "superfrank", "date": "2026-08-06", "stance": "warn", "quote": "they're incredibly inconsistent and the variance in output quality is much higher", "excerpt": "I've come to the same conclusion over and over with all of the Chinese models that have been claimed to be catching up with OpenAI's and Anthropic's frontier models (Deepseek 4, GLM 5.2, Kimi K3). At their best, I think they're closing in on Opus and GPT, but they're incredibly inconsistent and the variance in output quality is much higher than the best from any of the Anthropic or OpenAI models from the last few gen", "context": "", "use_case": "", "switched_from": "", "story": "Qwen3.8 Max now ranked as the best overall model by agentic index"}]}, {"label": "Slow responses", "summary": "It can be slow and take a long time to think, especially after the K3 launch.", "quotes": [{"id": 784000848, "item_id": 49000053, "url": "https://news.ycombinator.com/item?id=49000053", "by": "villish", "date": "2026-07-21", "stance": "recommend", "quote": "It's a good model otherwise.", "excerpt": "I'm not sure why you're so dismissive about this model. It isn't currently open, but it will be. It was slow when I used it so i'll give you that. There will be other providers with better performance. My only gripe was that it takes a very long time to think. It's a good model otherwise.", "context": "", "use_case": "", "switched_from": "", "story": "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"}, {"id": 785650832, "item_id": 49103177, "url": "https://news.ycombinator.com/item?id=49103177", "by": "zorked", "date": "2026-07-29", "stance": "recommend", "quote": "I have been using Kimi for months.", "excerpt": "I have been using Kimi for months. It did get superslow after the K3 launch, then they blocked signups and normalized the service. It's not ideal that people can't join but as a user I am happy that they focused on serving existing customers.", "context": "existing subscriber", "use_case": "AI model subscription", "switched_from": "", "story": "Kimi K3-256k"}]}, {"label": "Data training", "summary": "One commenter warns that the default terms allow training on user data.", "quotes": [{"id": 787164176, "item_id": 49197761, "url": "https://news.ycombinator.com/item?id=49197761", "by": "refulgentis", "date": "2026-08-06", "stance": "warn", "quote": "Their EULA is \u201cwe use your data and train on it unless you negotiate something with us privately\u201d", "excerpt": "Are you using Kimi w/expectation of ZDR? Their EULA is \u201cwe use your data and train on it unless you negotiate something with us privately\u201d", "context": "", "use_case": "", "switched_from": "", "story": "DeepSeek planning to significantly raise prices"}]}, {"label": "Usage caps", "summary": "The $19 plan's weekly cap was hit in a day and a half.", "quotes": [{"id": 784992032, "item_id": 49062002, "url": "https://news.ycombinator.com/item?id=49062002", "by": "Stagnant", "date": "2026-07-26", "stance": "recommend", "quote": "it is the first of the chinese models that actually feels comparable to Opus-tier models", "excerpt": "Yeah can't recommend their $19 plan, only took me a day and a half to hit the weekly usage cap. The $39 plan has 5x the limits so I recommend getting that instead. Having now tested it for a few days, it is the first of the chinese models that actually feels comparable to Opus-tier models", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Kimi K3 is not cheap"}]}], "who": [{"by": "rglover", "context": "does human-in-the-loop work", "url": "https://news.ycombinator.com/item?id=50242349", "date": "2026-10-26"}, {"by": "jackbeanstalk20", "context": "company planning to switch", "url": "https://news.ycombinator.com/item?id=50218495", "date": "2026-10-24"}, {"by": "jerieljan", "context": "moved to $100-tier subscription", "url": "https://news.ycombinator.com/item?id=50014997", "date": "2026-10-09"}, {"by": "abalashov", "context": "individual contributor", "url": "https://news.ycombinator.com/item?id=49815221", "date": "2026-09-23"}, {"by": "mark_l_watson", "context": "retired, former Google employee", "url": "https://news.ycombinator.com/item?id=49549353", "date": "2026-09-03"}, {"by": "pimeys", "context": "heavy work user of SOTA models", "url": "https://news.ycombinator.com/item?id=49407637", "date": "2026-08-23"}, {"by": "kromokromo", "context": "PM doing mostly knowledge work", "url": "https://news.ycombinator.com/item?id=49394040", "date": "2026-08-21"}, {"by": "SwellJoe", "context": "building a harness for security work", "url": "https://news.ycombinator.com/item?id=49295164", "date": "2026-08-13"}, {"by": "syntaxing", "context": "runs a Strix Halo server", "url": "https://news.ycombinator.com/item?id=49151061", "date": "2026-08-02"}, {"by": "__jl__", "context": "runs a production app", "url": "https://news.ycombinator.com/item?id=49113390", "date": "2026-07-30"}, {"by": "zorked", "context": "existing subscriber", "url": "https://news.ycombinator.com/item?id=49103177", "date": "2026-07-29"}, {"by": "martinjc", "context": "uses Claude at employer, Chinese models at home", "url": "https://news.ycombinator.com/item?id=49010201", "date": "2026-07-22"}], "tags": ["for coding", "for security research", "for plan reviews", "open weights", "for Rust dev"], "use_cases": ["coding", "ai coding", "coding tasks", "api-based coding", "coding team", "auditing code"], "compared": [{"slug": "glm", "name": "GLM", "co": 73, "switched_from": 2, "switched_to": 4}, {"slug": "deepseek", "name": "DeepSeek", "co": 46, "switched_from": 0, "switched_to": 2}, {"slug": "claude-opus", "name": "Claude Opus", "co": 43, "switched_from": 1, "switched_to": 0}, {"slug": "claude", "name": "Claude", "co": 30, "switched_from": 6, "switched_to": 0}, {"slug": "opencode", "name": "OpenCode", "co": 35, "switched_from": 0, "switched_to": 0}, {"slug": "claude-code", "name": "Claude Code", "co": 22, "switched_from": 2, "switched_to": 0}, {"slug": "deepseek-v4-flash", "name": "DeepSeek V4 Flash", "co": 20, "switched_from": 1, "switched_to": 2}, {"slug": "gpt-5", "name": "GPT-5", "co": 21, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=48999291", "title": "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA", "mentions": 7}, {"url": "https://news.ycombinator.com/item?id=48502347", "title": "Kimi K2.7-Code: open-source coding model with better token efficiency", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=50212065", "title": "Kimi K3.1 Pro and K3.1 Pro Highspeed", "mentions": 5}, {"url": "https://news.ycombinator.com/item?id=48935342", "title": "Kimi K3 is now live", "mentions": 5}, {"url": "https://news.ycombinator.com/item?id=49294997", "title": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities", "mentions": 4}, {"url": "https://news.ycombinator.com/item?id=49214008", "title": "DeepSeek V4 Flash 0731", "mentions": 4}], "sentiment_by_month": {"2025-11": {"recommend": 2, "neutral": 1, "compare": 1}, "2025-12": {"recommend": 1, "compare": 1}, "2026-01": {"neutral": 1, "compare": 1, "recommend": 1}, "2026-02": {"warn": 1, "recommend": 1}, "2026-03": {"recommend": 4}, "2026-04": {"recommend": 1, "neutral": 2}, "2026-05": {"recommend": 9, "compare": 1, "warn": 2, "neutral": 2}, "2026-06": {"recommend": 16, "neutral": 8, "compare": 4, "warn": 1}, "2026-07": {"compare": 5, "recommend": 22, "warn": 5, "neutral": 16}, "2026-08": {"neutral": 9, "recommend": 16, "warn": 7, "compare": 7}, "2026-09": {"compare": 4, "recommend": 7, "neutral": 4, "warn": 1}, "2026-10": {"recommend": 8, "neutral": 3, "warn": 2, "compare": 3}}, "quotes": [{"id": 804024531, "item_id": 50251533, "url": "https://news.ycombinator.com/item?id=50251533", "by": "lousken", "date": "2026-10-26", "stance": "recommend", "quote": "while GLM, Kimi and others stay the same", "excerpt": "My issue with anthropic and openai is that if I use a model like e.g. Sonnet 5 I don't always get full Sonnet 5, sometimes it's dumber, sometimes it is smarter (mostly dumber though), while GLM, Kimi and others stay the same. So if I rely on AI for things I better make sure the platform is stable and does not introduce hidden changes. Why is opus 5.5 suddenly a lot dumber? Who knows.", "context": "", "use_case": "", "switched_from": "", "story": "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"}, {"id": 803877584, "item_id": 50242349, "url": "https://news.ycombinator.com/item?id=50242349", "by": "rglover", "date": "2026-10-26", "stance": "recommend", "quote": "Now it's my daily driver and I have zero complaints.", "excerpt": "\"Cheap\" is up for debate, but the models themselves are great. Kimi K3 (and now 3.1 Pro) was the first open model I played with and didn't think to switch back to a frontier model. Now it's my daily driver and I have zero complaints. The frontier labs are in a tough spot. It's hard to justify their cost when you can get something that's easily as good for cheaper (and anecdotally, more reliable/accessible). I do HITL", "context": "does human-in-the-loop work", "use_case": "daily driver for coding", "switched_from": "", "story": "IBM CEO says open weight models are now good and cheap on-prem"}, {"id": 803871248, "item_id": 50241953, "url": "https://news.ycombinator.com/item?id=50241953", "by": "pwython", "date": "2026-10-26", "stance": "neutral", "quote": "I used Kimi K3.1 Pro to figure it out", "excerpt": "I used Kimi K3.1 Pro to figure it out as soon as cloudlandsdev put php-version.php live. The exploit worked just like this in a local Docker lab. But the terms page of the contest prohibited \"resource exhaustion,\" so it stopped there. Showing Kimi the PoC, it says: \"The winning chain deliberately used per-request memory exhaustion, which appears to violate the literal no-resource-exhaustion rule, although the organiz", "context": "", "use_case": "analyzing exploit", "switched_from": "", "story": "Steal My Bitcoin CTF PoC"}, {"id": 803831107, "item_id": 50239444, "url": "https://news.ycombinator.com/item?id=50239444", "by": "qlte", "date": "2026-10-26", "stance": "recommend", "quote": "I am entirely satisfied when using my Opencode Go plan with open models like Deepseek Flash/Pro, Kimi and MiMo.", "excerpt": "The profit comes from the enterprise customers who have to pay for usage at API rates. Not the subsidized subscriptions, which both OpenAI have designed to be unavailable for large enterprises by withholding IAM and auditing features they need except on the usage based entperise plans. As an individual I'm still paying for a ChatGPT Plus plan since $20/mo gives me a lot of value and I use many multiples of $20 at API", "context": "", "use_case": "general AI tasks", "switched_from": "", "story": "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"}, {"id": 803830656, "item_id": 50239416, "url": "https://news.ycombinator.com/item?id=50239416", "by": "ipsum2", "date": "2026-10-26", "stance": "warn", "quote": "Kimi is by far the most expensive in terms of tokens and $/work done.", "excerpt": "> much cheaper to access I'm calling BS. I use Kimi, chatGPT (Astra/Sol), and Claude (Opus 5.5), and Kimi is by far the most expensive in terms of tokens and $/work done. Using OpenCode harness.", "context": "", "use_case": "coding work", "switched_from": "", "story": "OpenAI and Anthropic cut their LLM prices by up to 80% within a few months. Why?"}, {"id": 803562656, "item_id": 50222666, "url": "https://news.ycombinator.com/item?id=50222666", "by": "chrisinnes", "date": "2026-10-24", "stance": "compare", "quote": "Found it functional but boring. Preferred Qwen.", "excerpt": "I used Kimi 2 (or 2.5) with my Openclaw and enjoyed it there so took out a subscription to Kimi 3.0 when it came out. Found it functional but boring. Preferred Qwen. Am not going to try the latest Kimi but it seems to be giving patchy results.", "context": "", "use_case": "LLM with OpenClaw", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 803556582, "item_id": 50222286, "url": "https://news.ycombinator.com/item?id=50222286", "by": "potsandpans", "date": "2026-10-24", "stance": "compare", "quote": "glm and kimi k.3, expensive, slow, ultimately not that much better than deepseek flash for this task.", "excerpt": "I recently just crossed the threshold for 100% decompilation of a PlayStation game. I used a lot of models in the process, and I feel like this is a good demonstration the \"sauce\" that opus has. The harness I've built for it is pretty straight forward: split binary into individual functions, give the agent a bunch of small tools to diff output with underlying binary, make them wrote clean c until there's a certified ", "context": "decompiling a PlayStation game", "use_case": "long-context compiler research", "switched_from": "", "story": "Kimi K3.1 scores 55 on Artificial Analysis, placed on the Pareto line"}, {"id": 803495920, "item_id": 50218495, "url": "https://news.ycombinator.com/item?id=50218495", "by": "jackbeanstalk20", "date": "2026-10-24", "stance": "recommend", "quote": "now this is just a better model with the same price", "excerpt": "I think my company will switch to this soon. We already switched from openAI to GLM and now this is just a better model with the same price", "context": "company planning to switch", "use_case": "company LLM", "switched_from": "GLM", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 803494432, "item_id": 50218402, "url": "https://news.ycombinator.com/item?id=50218402", "by": "sail0rm00n", "date": "2026-10-24", "stance": "compare", "quote": "Still burning through the 5-hour rate limit way too quickly, though. Output quality seems solid enough", "excerpt": "As tribute, I bit the bullet on the $40 plan to test out K3.1 Pro. Still burning through the 5-hour rate limit way too quickly, though. Output quality seems solid enough, but the value will probably make a lot more sense via API pricing for smaller projects. My current tier list based on what I\u2019ve been testing: 1. Opus 5.5 2. Astra 3. Argon 4. Kimi K3.1 Pro (For lightweight/daily driver tasks, Gemini Flash 3.8, DeepS", "context": "", "use_case": "coding", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 803469888, "item_id": 50216868, "url": "https://news.ycombinator.com/item?id=50216868", "by": "simondotau", "date": "2026-10-23", "stance": "recommend", "quote": "Kimi K3 has been my go-to model for its writing style", "excerpt": "Kimi K3 has been my go-to model for its writing style and (prior to Muse Wonder) my go-to for a third opinion during multi-round/multi-agent analysis tasks. Its writing style is IMHO so good that, for code which humans need to review, I regularly got it to clean up the painfully verbose comments left by Opus. I'm excited to see if its writing style and unique takes have been retained in this new release.", "context": "", "use_case": "writing and multi-agent analysis", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 803445840, "item_id": 50215365, "url": "https://news.ycombinator.com/item?id=50215365", "by": "3princip", "date": "2026-10-23", "stance": "warn", "quote": "Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "excerpt": "Same plan, just asked it to review a repo that Fable has been helping out with (max plan, haven't run out of the hourly or weekly limit once yet, and I use the same sub on multiple projects at once). Kimi didn't even finish the scan, ran out of the 5hr budget in less than 10 minutes. Unusable, unfortunately.", "context": "", "use_case": "reviewing a repo", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 802264432, "item_id": 50141527, "url": "https://news.ycombinator.com/item?id=50141527", "by": "verdverm", "date": "2026-10-18", "stance": "recommend", "quote": "models like Kimi K3 are right there, open and free to use, as an alternative to the closed models", "excerpt": "> I think it's still too early ... The data shows a clear trend, China is also doing the self improvement. From what we can tell, they are making the greater efficiency gains out of necessity (lack of easy access to the best hardware). Distillation is less important and models like Kimi K3 are right there, open and free to use, as an alternative to the closed models. I expect open weights to bypass closed models and ", "context": "", "use_case": "open-weights alternative to closed models", "switched_from": "", "story": "How the AI boom might end: 5 scenarios"}, {"id": 801861634, "item_id": 50116352, "url": "https://news.ycombinator.com/item?id=50116352", "by": "matthiasnehlsen", "date": "2026-10-16", "stance": "neutral", "quote": "compile the prompt for the agentic coding tool using Kimi K3", "excerpt": "Lotti is an open-source app for journaling, audio notes, photos, tasks and time tracking, synced end-to-end encrypted between your devices via Matrix. Or, if you prefer enterprise lingo, a distributed and open-source system of record for your memories. Lotti has an agentic layer where you get to pick where inference runs, including on your own device if powerful enough. Out on Linux via Flathub now (7500 downloads to", "context": "", "use_case": "compiling coding prompts", "switched_from": "", "story": "Lotti \u2013 Your memories, only on your devices, end-to-end encrypted"}, {"id": 801820737, "item_id": 50113796, "url": "https://news.ycombinator.com/item?id=50113796", "by": "verdverm", "date": "2026-10-16", "stance": "neutral", "quote": "1B on the $40 plan where I used more Kimi k3", "excerpt": "if you use mostly flash models, it goes on forever either way, the quotas are very generous, my usage pattern gets me around 250M tokens on the $10 plan, 1B on the $40 plan where I used more Kimi k3", "context": "", "use_case": "token usage", "switched_from": "", "story": "Ask HN: Thoughts about the Gemini Web app?"}, {"id": 800239952, "item_id": 50014997, "url": "https://news.ycombinator.com/item?id=50014997", "by": "jerieljan", "date": "2026-10-09", "stance": "recommend", "quote": "Kimi K2.5, 2.7 and Deepseek v4 were among my favorites", "excerpt": "I've been on the API-only mentality for months and all the open models were definitely the stuff I loved the most. Kimi K2.5, 2.7 and Deepseek v4 were among my favorites, while sparingly using Opus or whatever OpenAI had for specific situations. But ever since I've switched to one of the $100-tier subs, I can see why a lot of the people on it don't really discuss the open models often. I'd still use it especially whe", "context": "moved to $100-tier subscription", "use_case": "API-based coding", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 800239953, "item_id": 50014997, "url": "https://news.ycombinator.com/item?id=50014997", "by": "jerieljan", "date": "2026-10-09", "stance": "recommend", "quote": "Kimi K2.5, 2.7 and Deepseek v4 were among my favorites", "excerpt": "I've been on the API-only mentality for months and all the open models were definitely the stuff I loved the most. Kimi K2.5, 2.7 and Deepseek v4 were among my favorites, while sparingly using Opus or whatever OpenAI had for specific situations. But ever since I've switched to one of the $100-tier subs, I can see why a lot of the people on it don't really discuss the open models often. I'd still use it especially whe", "context": "", "use_case": "API-based coding", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 800181554, "item_id": 50011347, "url": "https://news.ycombinator.com/item?id=50011347", "by": "lmf4lol", "date": "2026-10-08", "stance": "recommend", "quote": "GLM-5.3 for coding or Kimi K.3 for research and critique.", "excerpt": "Oh man. v4.1-flash has been an sbolute game changer for us. We run all our Personal Assistants now on flash (thinking high) by default and it works incredibly well. There is really no need for basic agentic tasks that might require Kimi K.3 or GLM-5.3 levels. Once its gets juicier, we let flash launch specialized subagents with specific models. GLM-5.3 for coding or Kimi K.3 for research and critique. But as a main d", "context": "", "use_case": "research and critique", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 799665718, "item_id": 49979107, "url": "https://news.ycombinator.com/item?id=49979107", "by": "segmondy", "date": "2026-10-06", "stance": "neutral", "quote": "I use MiMov2.6Pro, DeepSeekv4.1Flash, GLM5.3, Hy4, Qwen3.8 and KimiK3 at home.", "excerpt": "I use Opus 5.5 at work. I use MiMov2.6Pro, DeepSeekv4.1Flash, GLM5.3, Hy4, Qwen3.8 and KimiK3 at home. Opus5.5 is not a game changer.", "context": "", "use_case": "home use", "switched_from": "", "story": "Mistral Large 4"}, {"id": 798997969, "item_id": 49937373, "url": "https://news.ycombinator.com/item?id=49937373", "by": "verdverm", "date": "2026-10-02", "stance": "recommend", "quote": "qwen3.8, kimi3, kimi2.7, GLM-5.3 are all good families I use in my coding team", "excerpt": "qwen3.8, kimi3, kimi2.7, GLM-5.3 are all good families I use in my coding team I'm mainly using flash varients, at least as the default, bump.up to stronger model as needed (less often these days)", "context": "", "use_case": "coding team", "switched_from": "", "story": "One month coding with GLM 5.3 Flash"}, {"id": 798997970, "item_id": 49937373, "url": "https://news.ycombinator.com/item?id=49937373", "by": "verdverm", "date": "2026-10-02", "stance": "recommend", "quote": "qwen3.8, kimi3, kimi2.7, GLM-5.3 are all good families I use in my coding team", "excerpt": "qwen3.8, kimi3, kimi2.7, GLM-5.3 are all good families I use in my coding team I'm mainly using flash varients, at least as the default, bump.up to stronger model as needed (less often these days)", "context": "", "use_case": "coding team", "switched_from": "", "story": "One month coding with GLM 5.3 Flash"}, {"id": 798302643, "item_id": 49893915, "url": "https://news.ycombinator.com/item?id=49893915", "by": "dools", "date": "2026-09-29", "stance": "warn", "quote": "since 2.7 Kimi has gone way downhill", "excerpt": "I go for weeks without topping up my Deepseek balance, and have to top up my OpenAI balance at least every week. GPT 6 Luna closed the gap significantly for sure (it seems to be about twice as expensive as DS v4.1 Flash), but Deepseek v4.1 Flash is still the best value model and capable enough for almost everything I need to do. Sometimes if it's babbling or can't nail down a solution I switch to Sol for one prompt, ", "context": "formerly a heavy Kimi k2.5/2.6 user", "use_case": "", "switched_from": "", "story": "Sonnet 5.5"}, {"id": 797326064, "item_id": 49832879, "url": "https://news.ycombinator.com/item?id=49832879", "by": "ben_w", "date": "2026-09-24", "stance": "neutral", "quote": "this was with Kimi which is one of the open models", "excerpt": "> It's unlikely that this is some AGI that spawned itself out of nothing and started doing this. You've not spent much time playing with these models, I see. Does't matter if you call the current models \"AGI\" or not, they: (1) successfully do stuff like this. Someone I know on Telegram found \"fifty or so\" Linux filesystems kernel bugs a few days ago, of which 26 were the first night while he slept; this was with Kimi", "context": "", "use_case": "finding Linux kernel bugs", "switched_from": "", "story": "OpenAI agent hacked Australian government website, PM says"}, {"id": 797043536, "item_id": 49815221, "url": "https://news.ycombinator.com/item?id=49815221", "by": "abalashov", "date": "2026-09-23", "stance": "recommend", "quote": "I would generally consider the SWE capabilities of Kimi to be on par with Claude", "excerpt": "Data points from my own usage, for whatever they're worth[1]. I'm just an individual contributor and hardly represent enterprise or large orgs: I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! 1) For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit i", "context": "individual contributor", "use_case": "everyday coding", "switched_from": "", "story": "The current balance of power in open models"}, {"id": 796953680, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "recommend", "quote": "The Chinese models are simply _excellent_", "excerpt": "Welcome to the dark side. I've been on Kimi, with a little DeepSeek-V4-Pro, GLM 5.2/5.3, and MiMo thrown in, for probably about a year now. It's great here! For DeepSeek, I recommend their Reasonix harness strongly, due to its alignment to DeepSeek's prefix cache. It means mostly (95%+) cache hit input tokens, so very cheap large-scale code analyses and things that require mega context windows (at the cost of some at", "context": "", "use_case": "everyday coding via OpenRouter", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}, {"id": 796706483, "item_id": 49794155, "url": "https://news.ycombinator.com/item?id=49794155", "by": "joshheitzman", "date": "2026-09-21", "stance": "neutral", "quote": "Kimi K2.5 before that", "excerpt": "Absolutely! DeepSeek-V4-Flash-0731 has become my daily driver. It's pretty amazing what it can do for what it costs at deepinfra.com (I don't use deepseek as a provider since they train on your data [at least their honest about it]). GLM-5.1 was my daily driver before that and Kimi K2.5 before that.", "context": "", "use_case": "daily driver model", "switched_from": "", "story": "MiMo v2.6"}, {"id": 796654657, "item_id": 49790916, "url": "https://news.ycombinator.com/item?id=49790916", "by": "eknkc", "date": "2026-09-21", "stance": "neutral", "quote": "Kimi and GLM feel extremely coding oriented. I use them for code reviews basically.", "excerpt": "I am subscribed to ChatGPT, Claude, Kimi and GLM coding plans. 200$ one on GPT and the 20$ ish ones on all others. Recently added Grok and it has somehow bacome my second most used model. For daily one off questions I prefer it because it is fast enough and I like the way it responds. I also use it for basic research like \u201cfind me a battery drill for this and that\u201d. Kimi and GLM feel extremely coding oriented. I use ", "context": "", "use_case": "code reviews", "switched_from": "", "story": "Grok 4.7"}, {"id": 795951680, "item_id": 49746980, "url": "https://news.ycombinator.com/item?id=49746980", "by": "arnav714412", "date": "2026-09-17", "stance": "recommend", "quote": "Kimmi with whisper.", "excerpt": "Kimmi with whisper. Ngl running it on my old laptop is the only thing keeping my thing from breaking apart", "context": "", "use_case": "self-hosted on old laptop", "switched_from": "", "story": "Ask HN: What self-hosted apps do you use daily?"}, {"id": 795685537, "item_id": 49730346, "url": "https://news.ycombinator.com/item?id=49730346", "by": "pimeys", "date": "2026-09-16", "stance": "compare", "quote": "For a long time Kimi K3 was the best model for this work", "excerpt": "My experience is complete opposite from yours. The V4 Flash was already quite good, but V4.1 is really very close to SOTA in my books. I've eval'd these models for weeks against Gemini 3.8, Kimi K3 and Opus 5, and DeepSeek absolutely wins these evals. It's really crazy. We price per token our customers. If Kimi was about 30% of the price of Opus 5 for the same quality, DeepSeek is 1/10th of a price of Kimi K3. We've ", "context": "", "use_case": "agentic coding work", "switched_from": "", "story": "DeepSeek v4.1 Flash Is Now Our Best Hacking Model"}, {"id": 795572705, "item_id": 49723294, "url": "https://news.ycombinator.com/item?id=49723294", "by": "weee322", "date": "2026-09-16", "stance": "recommend", "quote": "or kimi 3 on CPU", "excerpt": "Ideal for https://github.com/giannisanni/pulsar/ or kimi 3 on CPU probably 30s for one token is too long for normal people but for agents it is ok", "context": "", "use_case": "running agents on CPU", "switched_from": "", "story": "Show HN: Pizza Bot \u2013 An inbox for AI agents that work in the background"}, {"id": 795254548, "item_id": 49703409, "url": "https://news.ycombinator.com/item?id=49703409", "by": "criley2", "date": "2026-09-14", "stance": "compare", "quote": "offering Kimi k3 performance for like half the price", "excerpt": "I have been writing an internal code review tool that is a bit maximalist. I created subagents for many internal domains and technologies we manage, with prompts focused on best practices, common problems, owasp guidelines, etc), a separate tier of wider band subagents (design, rollout, security/privacy), and a final agent at the top orchestrating and combining. I also use adversarial validator passes against all fin", "context": "", "use_case": "code review", "switched_from": "", "story": "GPT-5.6 Luna vs. GPT-6 Astra: Is a $1.20 Model Good Enough for Code Review?"}, {"id": 794997650, "item_id": 49687353, "url": "https://news.ycombinator.com/item?id=49687353", "by": "aleda145", "date": "2026-09-13", "stance": "recommend", "quote": "vs Kimi 2.7 that I've been using in the cloud version", "excerpt": "I open sourced kavla, the SQL canvas I'm working on: https://github.com/aleda145/kavla README is very bare bones, needs a video or a picture too. The website should tell you some more things: https://kavla.dev (and now the demo does not require a login (also need to remake the demo to fit the new self hosted version)) It's still lacking an analytics agent for the local version though. I've explored using codex CLI, b", "context": "", "use_case": "analytics agent", "switched_from": "", "story": "Ask HN: What are you working on? (September 2026)"}, {"id": 794800336, "item_id": 49675021, "url": "https://news.ycombinator.com/item?id=49675021", "by": "abalashov", "date": "2026-09-12", "stance": "recommend", "quote": "Allow me to recommend switching to Kimi K3 / GLM 5.3 / DeepSeek.", "excerpt": "Allow me to recommend switching to Kimi K3 / GLM 5.3 / DeepSeek. I use all three for the better part of a year now (DeepSeek via the Reasonix harness) and haven't touched Claude Code in at least as long.", "context": "", "use_case": "coding", "switched_from": "Claude Code", "story": "We must pace the frontier"}, {"id": 794782419, "item_id": 49673901, "url": "https://news.ycombinator.com/item?id=49673901", "by": "gpugreg", "date": "2026-09-12", "stance": "neutral", "quote": "GLM-5.3, GLM-5.3-Flash and Kimi K3 are also fine, but slower, more expensive", "excerpt": "I currently use DeepSeek-V4.1-Flash and the V4-Pro and -Flash versions before that, because they are extremely cheap. I have spent less than $25 for over a billion token so far (1B cached, 15M out, 19M in). I even prefer the DeepSeek models over the older OpenAI offerings. When I tell a GPT-5.x model to do some difficult task, they often give up saying it can't be done, or cheat by modifying the tests, while the Deep", "context": "", "use_case": "programming", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 793031089, "item_id": 49564443, "url": "https://news.ycombinator.com/item?id=49564443", "by": "Roark66", "date": "2026-09-04", "stance": "compare", "quote": "even if I managed to switch to Kimi K3", "excerpt": "Don't they have very low limits? What do people use these tiny limits for? I started measuring my Claude Max x5 use and last week (they gave me 50% more) I used 1.3B input tokens. Some 130M were cache writes, rest was cached. And 5M output. This puts things in perspective. We're taking thousands of bucks weekly even if I managed to switch to Kimi K3. What is the majority of this use? Infrastructure upgrades, troubles", "context": "", "use_case": "", "switched_from": "", "story": "Qwen 3.8 27B available on Cerebras at 1500 tokens/s"}, {"id": 792973745, "item_id": 49560859, "url": "https://news.ycombinator.com/item?id=49560859", "by": "DaSHacka", "date": "2026-09-04", "stance": "recommend", "quote": "I just switch the model to Kimi K3 to finish up making the PoC", "excerpt": "Surprisingly, I've had really good luck with reverse engineering on frontier models (without being part of CVP or similar). It's more the exploit development/PoC that it locks up on, which, as I use `pi`, I just switch the model to Kimi K3 to finish up making the PoC. Ironically, due to the stringent guardrails on American models that exist to avoid giving adversaries a leg-up in cybersecurity, I end up feeding dozen", "context": "", "use_case": "exploit PoC development", "switched_from": "", "story": "GPT-6 Astra"}, {"id": 792789648, "item_id": 49549353, "url": "https://news.ycombinator.com/item?id=49549353", "by": "mark_l_watson", "date": "2026-09-03", "stance": "neutral", "quote": "I use something top-tier like Kimi K3", "excerpt": "good summary, thanks. I used to use Google Cloud for consulting and projects (I worked at Google for a while, and there is some nostalgia) so I have Gemini API via Google Cloud, but I am retired now. Your post reminded me that I need to shut that all down and switch to getting an API key using AI Studio. I have spent two months experimenting with a wide range of US and Chinese models, and I had a lot of fun doing tha", "context": "retired, former Google employee", "use_case": "difficult tasks occasionally", "switched_from": "", "story": "Gemini 3.8 Flash and 3.8 Flash Cyber"}, {"id": 792748850, "item_id": 49546803, "url": "https://news.ycombinator.com/item?id=49546803", "by": "nxdmum", "date": "2026-09-03", "stance": "recommend", "quote": "Combined with Kimi K3 for adversarial plan reviews - you get robust plans .", "excerpt": "so far using Gemini from 2.5 Pro to date (3.7 flash) - the way google trains the model it seems - is to identify top 3 to 4 things to fix first. as a result gemini is not that good in being thorough - but it's a needle mover . Opus5 Opus4.8 and Fable always point out things that Gemini missed. but Gemini was a needle mover - identifying the most important things to fix. I always enjoy interacting with Gemini . When i", "context": "", "use_case": "planning and plan review", "switched_from": "", "story": "Gemini 3.8 Flash and 3.8 Flash Cyber"}, {"id": 792748851, "item_id": 49546803, "url": "https://news.ycombinator.com/item?id=49546803", "by": "nxdmum", "date": "2026-09-03", "stance": "neutral", "quote": "Kimi 2.7 paired with Gemini 3.5/4.6/3.7 flash has been my implementer", "excerpt": "so far using Gemini from 2.5 Pro to date (3.7 flash) - the way google trains the model it seems - is to identify top 3 to 4 things to fix first. as a result gemini is not that good in being thorough - but it's a needle mover . Opus5 Opus4.8 and Fable always point out things that Gemini missed. but Gemini was a needle mover - identifying the most important things to fix. I always enjoy interacting with Gemini . When i", "context": "", "use_case": "implementation", "switched_from": "", "story": "Gemini 3.8 Flash and 3.8 Flash Cyber"}, {"id": 792584800, "item_id": 49536550, "url": "https://news.ycombinator.com/item?id=49536550", "by": "mark_l_watson", "date": "2026-09-02", "stance": "recommend", "quote": "I use Kimi K3 and GLM 5.3 to find old bugs and improve old code", "excerpt": "I use Kimi K3 and GLM 5.3 to find old bugs and improve old code, and would try Fable 5.1 if I could afford $50 for 1M output tokens (I am retired and all my research is self funded now) I like to start by prompting with \u201cexamine this code base for problems and improvements and write to IMPROVEMENTS.MD\u201d and then carefully look over the suggestions, and either fix myself or let the model+coding harness try.", "context": "retired, self-funded research", "use_case": "finding bugs, improving old code", "switched_from": "", "story": "Claude Fable 5.1 and Claude Mythos 5.1"}, {"id": 792546514, "item_id": 49534157, "url": "https://news.ycombinator.com/item?id=49534157", "by": "disgruntledphd2", "date": "2026-09-02", "stance": "compare", "quote": "I find it hard to distinguish between the outputs of GPT/Claude/Kimi/GLM recently", "excerpt": "Where does the secret sauce show up in the outputs, then? Like, (apart from tone), I find it hard to distinguish between the outputs of GPT/Claude/Kimi/GLM recently (I use cursor, and have been giving them the same prompt and comparing). If anything, I found that the non-Claude models were better in many cases, which definitely doesn't map to their pricing. > in fact not all the methods and data are public Probably n", "context": "", "use_case": "coding prompts", "switched_from": "", "story": "Path to Astra: critical capabilities and frontier safeguards"}, {"id": 790522192, "item_id": 49407637, "url": "https://news.ycombinator.com/item?id=49407637", "by": "pimeys", "date": "2026-08-23", "stance": "recommend", "quote": "Except a month ago I started testing Kimi K3 and omp and I never went back.", "excerpt": "I've been using all the SOTA models a lot at work, like serious amount of tokens. It's been really rare that I stick with one model and harness for too long... Except a month ago I started testing Kimi K3 and omp and I never went back. Something with this combo works really well for Rust dev. The model doesn't really annoy me at all and I have not switched to Opus or SOL. And the monthly token bill is much lower...", "context": "heavy work user of SOTA models", "use_case": "Rust development", "switched_from": "", "story": "A week of using Codex more than Claude"}, {"id": 789452800, "item_id": 49340800, "url": "https://news.ycombinator.com/item?id=49340800", "by": "Scene_Cast2", "date": "2026-08-17", "stance": "recommend", "quote": "Fwiw I love K3 and use it as a daily driver.", "excerpt": "Oh hey, that's cheaper than Kimi K3! Amusing to see a SOTA OpenAI model be cheaper than a Chinese open weight model. Fwiw I love K3 and use it as a daily driver. I haven't tried Sol, as I dislike OpenAI.", "context": "", "use_case": "daily driver", "switched_from": "", "story": "GPT-5.6 Sol Pricing Cut by 50%"}, {"id": 788722627, "item_id": 49295164, "url": "https://news.ycombinator.com/item?id=49295164", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "So, I switched to Kimi K3 and DeepSeek for API testing just because it's so much cheaper.", "excerpt": "I've been building a harness for security work, and had to switch to GPT 5.5 when even Opus started refusing security work. Then 5.6 Sol arrived, and it refuses security work, too. So, I switched to Kimi K3 and DeepSeek for API testing just because it's so much cheaper. But, if GLM is better, I'm here for it, as I think GLM is also cheaper than K3.", "context": "building a harness for security work", "use_case": "API testing", "switched_from": "GPT 5.6 Sol", "story": "GLM-5.3: Frontier Coding with Emergent Cyber Capabilities"}, {"id": 788624929, "item_id": 49289058, "url": "https://news.ycombinator.com/item?id=49289058", "by": "SwellJoe", "date": "2026-08-13", "stance": "recommend", "quote": "Kimi K3 is smarter, though.", "excerpt": "DeepSeek is more fun, because it's cheap-as-free, Good Enough, quite fast. I use it for all API stuff, automated runs, testing of the security auditing harness and benchmarks I'm working on, etc. Kimi K3 is smarter, though. At least smarter than DeepSeek V4 Flash 0731. I haven't tried the new Pro version, but will this weekend when I'm working on my personal projects But, K3 has been what I've been using for the actu", "context": "building a security auditing harness and benchmarks", "use_case": "coding a security auditing harness", "switched_from": "Claude", "story": "Qwen/Qwen3.8-2.4T-A95B"}, {"id": 788546979, "item_id": 49284186, "url": "https://news.ycombinator.com/item?id=49284186", "by": "gboss", "date": "2026-08-13", "stance": "warn", "quote": "Kimi 3 was super expensive as spec model", "excerpt": "For a personal project I\u2019ve been piloting Spec driven development (SDD), (it\u2019s contagious!), using Cursor and EARS statements. The strategy has been to use the frontier model to write the spec and a lessor model to write the tests and code and the frontier model to write critiquing prompts until it has nothing left to say. For my particular project there are two programs (or in human speak phases), where each program", "context": "", "use_case": "spec writing", "switched_from": "", "story": "Grok 4.6"}, {"id": 787474369, "item_id": 49217148, "url": "https://news.ycombinator.com/item?id=49217148", "by": "dools", "date": "2026-08-07", "stance": "warn", "quote": "I was using Kimi a lot but it just nose dived.", "excerpt": "I have been using deepseek v4 pro almost exclusively. I was using Kimi a lot but it just nose dived. The decline started with the release of 2.7 and accelerated with the release of 3. When I need vision capabilities I use GPT 5.3 codex and if deepseek can\u2019t figure something out after a few goes I switch to GTP 5.5 or 5.6 (I\u2019ve been giving Terra first bite recently and it does pretty well, and have used Sol a couple o", "context": "", "use_case": "", "switched_from": "", "story": "DeepSeek V4 Flash 0731"}, {"id": 785468576, "item_id": 49091786, "url": "https://news.ycombinator.com/item?id=49091786", "by": "rglover", "date": "2026-07-28", "stance": "recommend", "quote": "Just tried K3 out for the first time today and it's a legitimate threat.", "excerpt": "Just tried K3 out for the first time today and it's a legitimate threat. Temporarily (maybe permanently) using it as my daily driver but it's wild how comparable it is to Opus 4.7/4.8 (what's been my go to for a bit now\u2014wrote a quick post on what I found today [1]). [1] https://graybearding.bearblog.dev/kimi-k3-is-insane/", "context": "", "use_case": "daily driver coding", "switched_from": "Opus", "story": "Kimi K3 Architecture Overview and Notes"}, {"id": 784992032, "item_id": 49062002, "url": "https://news.ycombinator.com/item?id=49062002", "by": "Stagnant", "date": "2026-07-26", "stance": "recommend", "quote": "it is the first of the chinese models that actually feels comparable to Opus-tier models", "excerpt": "Yeah can't recommend their $19 plan, only took me a day and a half to hit the weekly usage cap. The $39 plan has 5x the limits so I recommend getting that instead. Having now tested it for a few days, it is the first of the chinese models that actually feels comparable to Opus-tier models", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Kimi K3 is not cheap"}, {"id": 784869632, "item_id": 49054352, "url": "https://news.ycombinator.com/item?id=49054352", "by": "alexsmirnov", "date": "2026-07-25", "stance": "recommend", "quote": "No rejections, they did find the same problems on my project that I used for testing.", "excerpt": "I did run both kimi-k3 and glm-5.2 with Capital One vulnhunt [1]. No rejections, they did find the same problems on my project that I used for testing. gpt-5.6, gemini pro, and opus all rejected to follow. gpt even declined to edit skill files. [1] https://www.capitalone.com/tech/open-source/announcing-vulnhunter/", "context": "", "use_case": "vulnerability hunting", "switched_from": "", "story": "UK AISI / Caisi Preliminary Assessment of Kimi K3's Cyber Capabilities"}, {"id": 783994737, "item_id": 48999671, "url": "https://news.ycombinator.com/item?id=48999671", "by": "apatheticonion", "date": "2026-07-21", "stance": "recommend", "quote": "now Kimi K3 offers a great planning assistant for more advanced coding tasks", "excerpt": "I love the Chinese models. I use DeepSeek exclusively and now Kimi K3 offers a great planning assistant for more advanced coding tasks. DeepSeek v4 Flash is extremely fast and is able to handle pretty much anything I've thrown at it (I use mostly Rust, PSQL, Angular and Terraform). I self host Bifrost as my LLM gateway, though I wish LLM vendors would do monthly/daily automatic billing (like VPS providers do) rather ", "context": "", "use_case": "planning for coding", "switched_from": "", "story": "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"}, {"id": 783809504, "item_id": 48988094, "url": "https://news.ycombinator.com/item?id=48988094", "by": "mempko", "date": "2026-07-20", "stance": "recommend", "quote": "Switched to K3 and continued my work. And it worked really well for what I was doing.", "excerpt": "I can second this. I was using Fable for some research and fable said it wasn't going to do it and downgraded to opus. Switched to K3 and continued my work. And it worked really well for what I was doing. You can easily switch claude code to use K3, it took like 30 seconds.", "context": "", "use_case": "research, as Claude Code backend", "switched_from": "Fable", "story": "Kimi Work"}]}