{"slug": "deepseek", "name": "DeepSeek", "category": "ai", "category_name": "AI models and assistants", "description": "Chinese AI lab's open-weight large language models, offered via chat app and low-cost API.", "recommenders": 162, "commenters": 250, "warners": 13, "score": 155.496, "rising": 27.02, "first": "2025-11-05", "latest": "2026-10-31", "why": "Liked for very low cost and fast output, with quality close enough to frontier models for most coding work.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 187, "neutral": 75, "compare": 15, "warn": 14}, "spark": [2, 2, 0, 0, 2, 2, 16, 56, 41, 19, 20, 27], "listed": true, "loved": ["Low cost", "Good-enough quality"], "caveat": "Hard tasks", "short_quote": "I have been using DeepSeek since forever and it's so good I was able to write a compiler and native desktop applications with it.", "icon": true, "domain": "deepseek.com", "github_repo": null, "official_url": "https://www.deepseek.com", "pricing": "open source", "aliases": ["DeepSeek", "DeepSeek V4 Pro", "DeepSeek V4", "DeepSeek-V4-Pro", "DeepSeek 4", "DeepSeek v4", "DeepSeek V4-Pro", "Deepseek V4 pro"], "verdict": [{"text": "Liked for very low cost and fast output", "cites": [{"id": 804018448, "item_id": 50251153, "url": "https://news.ycombinator.com/item?id=50251153", "by": "dudisubekti", "date": "2026-10-26", "stance": "recommend", "quote": "It's dirt cheap, blazing fast, and is more than enough for most coding work.", "excerpt": "I use it at work. It's dirt cheap, blazing fast, and is more than enough for most coding work. If you live in a country with a weak currency, DeepSeek models are definitely a godsend.", "context": "", "use_case": "coding at work", "switched_from": "", "story": "GPT-6.1 Terra"}, {"id": 798373696, "item_id": 49898356, "url": "https://news.ycombinator.com/item?id=49898356", "by": "proxysna", "date": "2026-09-29", "stance": "recommend", "quote": "Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper", "excerpt": "I am yet to spend $200 on deepseek this year. Not sure what kind of usage can justify $200/month of either openai or anthropic, i'm not even talking about $500. Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper that i no longer care about how much i use it. I never hit any daily/weekly quota or anything like that while working or tinkering. At this point i am OK with being 6 months ", "context": "", "use_case": "everyday coding and tinkering", "switched_from": "", "story": "GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price"}, {"id": 796993488, "item_id": 49812093, "url": "https://news.ycombinator.com/item?id=49812093", "by": "knapcio", "date": "2026-09-23", "stance": "recommend", "quote": "Then I switched to DeepSeek and haven\u2019t noticed any drop in quality. It\u2019s fast at around 300 TPS and cheap.", "excerpt": "I\u2019ve been working with frontier models for the last year on a large real-world project with multiple apps. Eventually they started becoming unreliable. Simple UI issues, usage limits and overengineering became constant problems. I had to come up with a robust process: task analysis, user intent analysis, implementation and testing. All of these steps loop when needed with 15-20 steps per task in total. This finally g", "context": "works on a large real-world multi-app project", "use_case": "coding on large project", "switched_from": "", "story": "Claude Opus 5.5"}]}, {"text": ", with quality close enough to frontier models for most coding work.", "cites": [{"id": 803899730, "item_id": 50243733, "url": "https://news.ycombinator.com/item?id=50243733", "by": "Sphax", "date": "2026-10-26", "stance": "recommend", "quote": "These chinese models are simply good enough for the vast majority of tasks I run", "excerpt": "I use Opus 5.5 daily at work and a mix of GLM 5.3 (Flash or not), Deepseek v4 (flash or pro) at home and while Opus 5.5 is more capable, in my opinion it matters for less than 5% of the work i do. These chinese models are simply good enough for the vast majority of tasks I run and when they're not I use Astra or Opus 5.5. Yes, leaving Opus 5.5 on all the time is comfortable when you pay the 90\u20ac plan but it's just not", "context": "", "use_case": "home use", "switched_from": "", "story": "Anthropic S-1"}, {"id": 803950912, "item_id": 50246932, "url": "https://news.ycombinator.com/item?id=50246932", "by": "codebolt", "date": "2026-10-26", "stance": "recommend", "quote": "struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable", "excerpt": "I was playing around with OpenCode on Omarchy over the weekend, working on my Android app, and was struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable (where I typically rack up at least a few hundred dollars per month).", "context": "works on an Android app", "use_case": "coding Android app", "switched_from": "", "story": "Anthropic S-1"}, {"id": 802523680, "item_id": 50157730, "url": "https://news.ycombinator.com/item?id=50157730", "by": "surgical_fire", "date": "2026-10-19", "stance": "recommend", "quote": "DeepSeek probably could do everything I needed.", "excerpt": "> claude gives the edge I have no idea what you work with, but for every single programming job I had in more than 2 decades, from the lamest consultancy to FAANG, you don't need the Edge. DeepSeek probably could do everything I needed. Would I pay for LLMs? Sure, but not much more than 30-50 bucks range per month. I can affort 1k/m, but no way in hell I would accept it. I would rather just not use LLMs (it's not lik", "context": "20+ years programming", "use_case": "programming work", "switched_from": "", "story": "The obscure trust looming over Anthropic's IPO"}]}, {"text": " The most common caveat is that it trails top models like Opus on very hard tasks and UI design", "cites": [{"id": 802106800, "item_id": 50131675, "url": "https://news.ycombinator.com/item?id=50131675", "by": "dw_arthur", "date": "2026-10-17", "stance": "recommend", "quote": "I am a huge DeepSeek fan", "excerpt": "I am a huge DeepSeek fan, but it cannot come close to Opus 5.5 on UI design. A $20 Claude subscription that I use some months and cancel the others along with Deepseek API is great.", "context": "", "use_case": "coding via API", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}, {"id": 802560321, "item_id": 50160020, "url": "https://news.ycombinator.com/item?id=50160020", "by": "mlcruz", "date": "2026-10-20", "stance": "recommend", "quote": "Deepseek can do 90% of the job", "excerpt": "I think we keep getting used to intelligence improvements and the goal posts are always moving forward. I remember that close to an year ago i felt the same way about opus 4.5. And even if you can easily run ds or qwen locally to get very similar results to that, we moved to 5.5 being the new target. Deepseek can do 90% of the job but a claude reviwer/spec writer is a brutal improvement.", "context": "", "use_case": "coding", "switched_from": "", "story": "MiniMax M3.1 Flash"}]}, {"text": ", and some avoid it over data-privacy concerns.", "cites": [{"id": 796680674, "item_id": 49792542, "url": "https://news.ycombinator.com/item?id=49792542", "by": "RussianCow", "date": "2026-09-21", "stance": "warn", "quote": "The downside is they train their models on your input, which might be a deal-breaker for many users (as it is for me).", "excerpt": "For caching, only if you don't specify your preferred providers and let OpenRouter route each request itself. I have stuff like this in my OpenCode config for each model I use and I regularly get ~90-95% cache hit rates. \"order\": [\"relace\", \"coreweave\", \"novita\", \"baseten\", \"together\"], \"allow_fallbacks\": false It still won't be quite as high as you'd get by just using DeepSeek because occasionally a request will fai", "context": "", "use_case": "direct API use", "switched_from": "", "story": "Grok 4.7"}, {"id": 796953681, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "recommend", "quote": "very cheap large-scale code analyses and things that require mega context windows", "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": "large-scale code analysis", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}]}], "praise": [{"label": "Low cost", "summary": "Commenters call it dirt cheap and say they rarely spend much on it even with heavy use.", "quotes": [{"id": 804018448, "item_id": 50251153, "url": "https://news.ycombinator.com/item?id=50251153", "by": "dudisubekti", "date": "2026-10-26", "stance": "recommend", "quote": "It's dirt cheap, blazing fast, and is more than enough for most coding work.", "excerpt": "I use it at work. It's dirt cheap, blazing fast, and is more than enough for most coding work. If you live in a country with a weak currency, DeepSeek models are definitely a godsend.", "context": "", "use_case": "coding at work", "switched_from": "", "story": "GPT-6.1 Terra"}, {"id": 803950912, "item_id": 50246932, "url": "https://news.ycombinator.com/item?id=50246932", "by": "codebolt", "date": "2026-10-26", "stance": "recommend", "quote": "struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable", "excerpt": "I was playing around with OpenCode on Omarchy over the weekend, working on my Android app, and was struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable (where I typically rack up at least a few hundred dollars per month).", "context": "works on an Android app", "use_case": "coding Android app", "switched_from": "", "story": "Anthropic S-1"}, {"id": 793955074, "item_id": 49622192, "url": "https://news.ycombinator.com/item?id=49622192", "by": "comrade1234", "date": "2026-09-09", "stance": "recommend", "quote": "I stick with deepseek because it's more than good enough and is super cheap.", "excerpt": "Does it really? Or is it that they're all pretty much good enough and so you just go with what's cheapest or what you're used to? For example, in IntelliJ I can switch between maybe a dozen models with a drop-down but the only time I do it is when whatever model I'm currently using has an outage. Another example, our product has some ai features (not coding related). I use spring-ai and so I can change models just in", "context": "", "use_case": "AI features in product", "switched_from": "", "story": "AI Has a Discovery Problem"}]}, {"label": "Good-enough quality", "summary": "Many find the gap to frontier models small or unnoticeable for most of their tasks.", "quotes": [{"id": 803899730, "item_id": 50243733, "url": "https://news.ycombinator.com/item?id=50243733", "by": "Sphax", "date": "2026-10-26", "stance": "recommend", "quote": "These chinese models are simply good enough for the vast majority of tasks I run", "excerpt": "I use Opus 5.5 daily at work and a mix of GLM 5.3 (Flash or not), Deepseek v4 (flash or pro) at home and while Opus 5.5 is more capable, in my opinion it matters for less than 5% of the work i do. These chinese models are simply good enough for the vast majority of tasks I run and when they're not I use Astra or Opus 5.5. Yes, leaving Opus 5.5 on all the time is comfortable when you pay the 90\u20ac plan but it's just not", "context": "", "use_case": "home use", "switched_from": "", "story": "Anthropic S-1"}, {"id": 796993488, "item_id": 49812093, "url": "https://news.ycombinator.com/item?id=49812093", "by": "knapcio", "date": "2026-09-23", "stance": "recommend", "quote": "Then I switched to DeepSeek and haven\u2019t noticed any drop in quality. It\u2019s fast at around 300 TPS and cheap.", "excerpt": "I\u2019ve been working with frontier models for the last year on a large real-world project with multiple apps. Eventually they started becoming unreliable. Simple UI issues, usage limits and overengineering became constant problems. I had to come up with a robust process: task analysis, user intent analysis, implementation and testing. All of these steps loop when needed with 15-20 steps per task in total. This finally g", "context": "works on a large real-world multi-app project", "use_case": "coding on large project", "switched_from": "", "story": "Claude Opus 5.5"}, {"id": 802256304, "item_id": 50141019, "url": "https://news.ycombinator.com/item?id=50141019", "by": "pimeys", "date": "2026-10-18", "stance": "recommend", "quote": "now I just do planning with DeepSeek too and don't see much difference", "excerpt": "I also used to plan with more expensive models like you did, but now I just do planning with DeepSeek too and don't see much difference. I use the omp plan mode and iterate on the text until I'm happy with it. Then clear context, give the plan to DeepSeek again for implementation. If it's mission-critical code, do a security review with GLM 5.3. Otherwise go manually through the code and ask for changes if needed. Wo", "context": "", "use_case": "planning and implementation", "switched_from": "more expensive models", "story": "DeepSeek-v4.1-Flash-1017"}]}, {"label": "Speed", "summary": "Users describe it as fast, with one citing around 300 tokens per second.", "quotes": [{"id": 804018448, "item_id": 50251153, "url": "https://news.ycombinator.com/item?id=50251153", "by": "dudisubekti", "date": "2026-10-26", "stance": "recommend", "quote": "It's dirt cheap, blazing fast, and is more than enough for most coding work.", "excerpt": "I use it at work. It's dirt cheap, blazing fast, and is more than enough for most coding work. If you live in a country with a weak currency, DeepSeek models are definitely a godsend.", "context": "", "use_case": "coding at work", "switched_from": "", "story": "GPT-6.1 Terra"}, {"id": 798373696, "item_id": 49898356, "url": "https://news.ycombinator.com/item?id=49898356", "by": "proxysna", "date": "2026-09-29", "stance": "recommend", "quote": "Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper", "excerpt": "I am yet to spend $200 on deepseek this year. Not sure what kind of usage can justify $200/month of either openai or anthropic, i'm not even talking about $500. Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper that i no longer care about how much i use it. I never hit any daily/weekly quota or anything like that while working or tinkering. At this point i am OK with being 6 months ", "context": "", "use_case": "everyday coding and tinkering", "switched_from": "", "story": "GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price"}, {"id": 796993488, "item_id": 49812093, "url": "https://news.ycombinator.com/item?id=49812093", "by": "knapcio", "date": "2026-09-23", "stance": "recommend", "quote": "Then I switched to DeepSeek and haven\u2019t noticed any drop in quality. It\u2019s fast at around 300 TPS and cheap.", "excerpt": "I\u2019ve been working with frontier models for the last year on a large real-world project with multiple apps. Eventually they started becoming unreliable. Simple UI issues, usage limits and overengineering became constant problems. I had to come up with a robust process: task analysis, user intent analysis, implementation and testing. All of these steps loop when needed with 15-20 steps per task in total. This finally g", "context": "works on a large real-world multi-app project", "use_case": "coding on large project", "switched_from": "", "story": "Claude Opus 5.5"}]}, {"label": "Daily driver", "summary": "Several use it for nearly everything and say it executes consistently.", "quotes": [{"id": 802115936, "item_id": 50132246, "url": "https://news.ycombinator.com/item?id=50132246", "by": "sligbad", "date": "2026-10-17", "stance": "recommend", "quote": "daily driver since V4", "excerpt": "Once read, I think from G Huntley, \"pick a model and stick with it,\" they have been making that very easy, daily driver since V4.", "context": "", "use_case": "daily AI model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802069840, "item_id": 50129365, "url": "https://news.ycombinator.com/item?id=50129365", "by": "ytxm", "date": "2026-10-17", "stance": "recommend", "quote": "I use it for literally everything.", "excerpt": "It passed my threshold for not caring to try new models back with DS4, I use it for literally everything. Announcements of new models in the series always feel curious, because it doesn't really feel like there's much left to fix that would improve on what's already there for any task I know of. It's a bit like obsessing over patch releases of some software after a major upgrade already solved the original problem lo", "context": "", "use_case": "general-purpose AI model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}]}, {"label": "Local running", "summary": "Some run quantized versions on their own hardware for real work.", "quotes": [{"id": 802085440, "item_id": 50130340, "url": "https://news.ycombinator.com/item?id=50130340", "by": "davedx", "date": "2026-10-17", "stance": "recommend", "quote": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.", "excerpt": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well. It can one shot iOS apps on my Mac Studio, it's a little slow and sometimes needs retries but can definitely do medium difficulty real work for me. I didn't try the last flash model because I would need to use SSD streaming (I have 128 GB), but the benches for this release are so compelling (and it usin", "context": "runs local models on a Mac Studio with 128 GB", "use_case": "local coding model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 798462080, "item_id": 49903880, "url": "https://news.ycombinator.com/item?id=49903880", "by": "rpdillon", "date": "2026-09-30", "stance": "recommend", "quote": "run a two-bit quant of deepseek on my Strix Halo", "excerpt": "Local AI has completely changed my relationship with computers because now I can build utilities, tools, scripts, and get debugging help with my systems with far less friction than I ever could before. I used to do all of this by hand, but now I can spin it off as essentially an async sub-process to the AI that I can return to a few hours later and have the job done. Everything from system upgrades to custom Steam De", "context": "home lab user", "use_case": "local scripting and debugging help", "switched_from": "", "story": "AI needs $6T in annual revenue to justify data centre boom"}]}, {"label": "Fewer refusals", "summary": "Users note it helped with tasks other models refused and is available uncensored through some services.", "quotes": [{"id": 795688320, "item_id": 49730520, "url": "https://news.ycombinator.com/item?id=49730520", "by": "trollbridge", "date": "2026-09-16", "stance": "recommend", "quote": "I use DS straight off of DeepSeek's API.", "excerpt": "I use DS straight off of DeepSeek's API. It was very helpful in the aftermath of a client whose WordPress got stuffed up and analysing when it happened, how it happened, and what the entry point was. The American frontier models refused to help because, well, they just refuse to help with that kind of thing.", "context": "does work for clients", "use_case": "investigating a hacked WordPress site", "switched_from": "", "story": "DeepSeek v4.1 Flash Is Now Our Best Hacking Model"}, {"id": 804212720, "item_id": 50263295, "url": "https://news.ycombinator.com/item?id=50263295", "by": "ecommerceguy", "date": "2026-10-27", "stance": "recommend", "quote": "I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek", "excerpt": "I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek, and, quite a few others actually. Fun to use and cheap. I have no idea how anyone is going to profit off this stuff. I mean I make videos, images, analysis, presentations etc etc with this studio. $10 a month. Web interface. In addition to the 15 others I use randomly for generic searches.", "context": "", "use_case": "videos, images, analysis, presentations", "switched_from": "", "story": "There are now 5 inference providers for abliterated DeepSeek"}, {"id": 790145872, "item_id": 49384117, "url": "https://news.ycombinator.com/item?id=49384117", "by": "lemontheme", "date": "2026-08-20", "stance": "recommend", "quote": "deepseek v4 will happily discuss it", "excerpt": "Hah, I use the same probe when I\u2019m unsure which provider openrouter is routing me to! Fwiw, deepseek v4 will happily discuss it. Only chinese providers will stop it in its tracks and give a canned answer. Streamed responses sometimes start with what the model was actually generating before it got cut off. It\u2019s top bad, really. Sometimes the Chinese providers are the model labs themselves, like deepseek. I\u2019d like my m", "context": "", "use_case": "LLM via OpenRouter", "switched_from": "", "story": "Stealth Model"}]}], "caveats": [{"label": "Hard tasks", "summary": "It falls short of Opus on UI design and can fail consistently on very hard problems.", "quotes": [{"id": 802106800, "item_id": 50131675, "url": "https://news.ycombinator.com/item?id=50131675", "by": "dw_arthur", "date": "2026-10-17", "stance": "recommend", "quote": "I am a huge DeepSeek fan", "excerpt": "I am a huge DeepSeek fan, but it cannot come close to Opus 5.5 on UI design. A $20 Claude subscription that I use some months and cancel the others along with Deepseek API is great.", "context": "", "use_case": "coding via API", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}, {"id": 802560321, "item_id": 50160020, "url": "https://news.ycombinator.com/item?id=50160020", "by": "mlcruz", "date": "2026-10-20", "stance": "recommend", "quote": "Deepseek can do 90% of the job", "excerpt": "I think we keep getting used to intelligence improvements and the goal posts are always moving forward. I remember that close to an year ago i felt the same way about opus 4.5. And even if you can easily run ds or qwen locally to get very similar results to that, we moved to 5.5 being the new target. Deepseek can do 90% of the job but a claude reviwer/spec writer is a brutal improvement.", "context": "", "use_case": "coding", "switched_from": "", "story": "MiniMax M3.1 Flash"}]}, {"label": "Data privacy", "summary": "Commenters note training on inputs, data sent to China, and employers wary of Chinese models.", "quotes": [{"id": 796680674, "item_id": 49792542, "url": "https://news.ycombinator.com/item?id=49792542", "by": "RussianCow", "date": "2026-09-21", "stance": "warn", "quote": "The downside is they train their models on your input, which might be a deal-breaker for many users (as it is for me).", "excerpt": "For caching, only if you don't specify your preferred providers and let OpenRouter route each request itself. I have stuff like this in my OpenCode config for each model I use and I regularly get ~90-95% cache hit rates. \"order\": [\"relace\", \"coreweave\", \"novita\", \"baseten\", \"together\"], \"allow_fallbacks\": false It still won't be quite as high as you'd get by just using DeepSeek because occasionally a request will fai", "context": "", "use_case": "direct API use", "switched_from": "", "story": "Grok 4.7"}, {"id": 796953681, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "recommend", "quote": "very cheap large-scale code analyses and things that require mega context windows", "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": "large-scale code analysis", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}]}, {"label": "Context quirks", "summary": "It sometimes wrongly believes it is running out of context, and long contexts can cause attentional drift.", "quotes": [{"id": 798512000, "item_id": 49907000, "url": "https://news.ycombinator.com/item?id=49907000", "by": "rednb", "date": "2026-09-30", "stance": "recommend", "quote": "I use the direct API from deepseek using Opencode, no problem to report, works like a charm.", "excerpt": "I use the direct API from deepseek using Opencode, no problem to report, works like a charm. Except maybe that the model often believes that he is running out of context, and needs to rush so i occasionally need to jump in to tell it that it still has plenty of room left. But this does not degrade the quality of my overall experience in a meaningful way.", "context": "", "use_case": "coding via API", "switched_from": "", "story": "Livenerf: Has Opus 5.5 been nerfed yet?"}, {"id": 796953681, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "recommend", "quote": "very cheap large-scale code analyses and things that require mega context windows", "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": "large-scale code analysis", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}]}, {"label": "Mixed results", "summary": "One user got nothing useful from it, and another argues review time outweighs its API savings.", "quotes": [{"id": 797999235, "item_id": 49874952, "url": "https://news.ycombinator.com/item?id=49874952", "by": "sitsitstay", "date": "2026-09-28", "stance": "warn", "quote": "it did nothing good for me in the scheme of things", "excerpt": "I've been tokenmaxxing across Claude (multiple subs, but I've now burned all of them - and also exhausted, for all of them, the discounted \"Buy Credits\" from 40% to 25% to 20%....). I used Astra, and it lasted less than hour before it was dead for a month - and Grok whilst did solid work, struggled with cohesive architecture. DeepSeek, well, I've got 400+ dollars left in credit I will donate to my enemies, to slow th", "context": "", "use_case": "coding", "switched_from": "", "story": "Open Annihilation \u2013 Spent the GDP of a Small Nation Remaking a Classic 90s Game"}, {"id": 791441986, "item_id": 49465124, "url": "https://news.ycombinator.com/item?id=49465124", "by": "criley2", "date": "2026-08-27", "stance": "warn", "quote": "A single turn around in a PR costs more in human time than the difference between deepseek and fable in API costs.", "excerpt": "I'm sorry, but just because you achieve results you consider acceptable with this method doesn't mean everyone does. I don't work where we can ship slop. I don't work where PRs can be merged based on what the agents say. I work where a human has to read and approve and own every single line of code. I work where the stakes are actually high, so the cost of not using the best tools in terms of human time are big. A si", "context": "works where stakes are high and humans review every line", "use_case": "cheap agent implementation", "switched_from": "", "story": "The Harness Is the Thing"}]}, {"label": "Local speed", "summary": "Running it locally can be slow and sometimes needs retries.", "quotes": [{"id": 802085440, "item_id": 50130340, "url": "https://news.ycombinator.com/item?id=50130340", "by": "davedx", "date": "2026-10-17", "stance": "recommend", "quote": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.", "excerpt": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well. It can one shot iOS apps on my Mac Studio, it's a little slow and sometimes needs retries but can definitely do medium difficulty real work for me. I didn't try the last flash model because I would need to use SSD streaming (I have 128 GB), but the benches for this release are so compelling (and it usin", "context": "runs local models on a Mac Studio with 128 GB", "use_case": "local coding model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}]}, {"label": "Tooling jank", "summary": "Its harness with computer use and browser integration is described as somewhat janky.", "quotes": [{"id": 797473136, "item_id": 49842071, "url": "https://news.ycombinator.com/item?id=49842071", "by": "akmarinov", "date": "2026-09-25", "stance": "recommend", "quote": "It's been rather awesome", "excerpt": "I'm using the Deepseek harness with the deepseek computer use and deepseek browser integration, it's somewhat janky but it gets the job done. Also for coding i use OMP directly with Deepseek. It's been rather awesome because even maxed out $20 gets you very far in between Codex resets. I also tried developing a small game where I have Astra use as many Deepseek workers and it needs and it only does the 3D objects and", "context": "", "use_case": "coding and computer use", "switched_from": "", "story": "Best LLM for every budget, updated daily"}]}], "who": [{"by": "athrowaway3z", "context": "complete amateur", "url": "https://news.ycombinator.com/item?id=50288750", "date": "2026-10-29"}, {"by": "codebolt", "context": "works on an Android app", "url": "https://news.ycombinator.com/item?id=50246932", "date": "2026-10-26"}, {"by": "solomonb", "context": "homelab user", "url": "https://news.ycombinator.com/item?id=50169044", "date": "2026-10-20"}, {"by": "surgical_fire", "context": "20+ years programming", "url": "https://news.ycombinator.com/item?id=50157730", "date": "2026-10-19"}, {"by": "aizk", "context": "works for an open source AI inference provider", "url": "https://news.ycombinator.com/item?id=50135410", "date": "2026-10-17"}, {"by": "davedx", "context": "runs local models on a Mac Studio with 128 GB", "url": "https://news.ycombinator.com/item?id=50130340", "date": "2026-10-17"}, {"by": "denkmoon", "context": "uses frontier models at work", "url": "https://news.ycombinator.com/item?id=50013406", "date": "2026-10-08"}, {"by": "rpdillon", "context": "home lab user", "url": "https://news.ycombinator.com/item?id=49903880", "date": "2026-09-30"}, {"by": "switchbak", "context": "uses frontier models at work", "url": "https://news.ycombinator.com/item?id=49823054", "date": "2026-09-23"}, {"by": "abalashov", "context": "individual contributor", "url": "https://news.ycombinator.com/item?id=49815221", "date": "2026-09-23"}, {"by": "knapcio", "context": "works on a large real-world multi-app project", "url": "https://news.ycombinator.com/item?id=49812093", "date": "2026-09-23"}, {"by": "apatheticonion", "context": "recently started a job that only uses Claude models", "url": "https://news.ycombinator.com/item?id=49810699", "date": "2026-09-23"}], "tags": ["for coding", "for budget users", "runs locally", "for homelabs", "for side projects"], "use_cases": ["coding", "ai coding", "personal projects", "coding model", "coding via api", "coding assistance"], "compared": [{"slug": "claude", "name": "Claude", "co": 69, "switched_from": 14, "switched_to": 0}, {"slug": "glm", "name": "GLM", "co": 68, "switched_from": 2, "switched_to": 1}, {"slug": "opencode", "name": "OpenCode", "co": 68, "switched_from": 0, "switched_to": 1}, {"slug": "claude-opus", "name": "Claude Opus", "co": 49, "switched_from": 3, "switched_to": 0}, {"slug": "kimi", "name": "Kimi", "co": 46, "switched_from": 2, "switched_to": 0}, {"slug": "claude-code", "name": "Claude Code", "co": 35, "switched_from": 3, "switched_to": 0}, {"slug": "pi", "name": "Pi", "co": 35, "switched_from": 0, "switched_to": 0}, {"slug": "openrouter", "name": "OpenRouter", "co": 32, "switched_from": 1, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=50129045", "title": "DeepSeek-v4.1-Flash-1017", "mentions": 10}, {"url": "https://news.ycombinator.com/item?id=49119559", "title": "DeepSeek-V4-Flash Update", "mentions": 9}, {"url": "https://news.ycombinator.com/item?id=48413629", "title": "Ask HN: What is your (AI) dev tech stack / workflow? (June 2026)", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=49274600", "title": "DeepSeek V4 Pro 0813", "mentions": 5}, {"url": "https://news.ycombinator.com/item?id=49214008", "title": "DeepSeek V4 Flash 0731", "mentions": 5}, {"url": "https://news.ycombinator.com/item?id=48518969", "title": "AI Coding at Home Without Going Broke", "mentions": 5}], "sentiment_by_month": {"2025-11": {"recommend": 2}, "2025-12": {"warn": 1, "neutral": 1, "recommend": 2, "compare": 1}, "2026-01": {"neutral": 1}, "2026-02": {}, "2026-03": {"recommend": 2, "neutral": 1}, "2026-04": {"recommend": 2, "neutral": 2}, "2026-05": {"recommend": 16, "compare": 3, "neutral": 4, "warn": 2}, "2026-06": {"neutral": 21, "recommend": 56, "compare": 4, "warn": 1}, "2026-07": {"recommend": 41, "neutral": 14, "compare": 2, "warn": 1}, "2026-08": {"recommend": 19, "warn": 6, "neutral": 12, "compare": 1}, "2026-09": {"neutral": 10, "recommend": 20, "compare": 2, "warn": 2}, "2026-10": {"neutral": 9, "recommend": 27, "compare": 2, "warn": 1}}, "quotes": [{"id": 805051376, "item_id": 50315711, "url": "https://news.ycombinator.com/item?id=50315711", "by": "paper2d", "date": "2026-10-31", "stance": "recommend", "quote": "I used deepseek and pi harness and it decompiled it via ghidra and recompiled it for wasm in under 40 mins", "excerpt": "I did something similar with an rare, old 32bit game I really enjoyed in my childhood. I used deepseek and pi harness and it decompiled it via ghidra and recompiled it for wasm in under 40 mins or so and made it run in the browser. I was impressed! For anyone who wants to run the game in your browser, here it is - https://bonkheads-port-by-kapil.pages.dev/", "context": "", "use_case": "decompiling and porting old game", "switched_from": "", "story": "Cisco gives engineer unlimited AI tokens to rewrite legacy code"}, {"id": 804620000, "item_id": 50288750, "url": "https://news.ycombinator.com/item?id=50288750", "by": "athrowaway3z", "date": "2026-10-29", "stance": "neutral", "quote": "I'm a complete amateur that had DeepSeek build a PHP2Shell exploit", "excerpt": "> A1 \u2014 Memory Exhaustion (~383MB): I'm actually somewhat surprised at the clumsiness here. I'm a complete amateur that had DeepSeek build a PHP2Shell exploit, and once the stealmybitcoin site went up and set an upload limit, i had DeepSeek try to shrink/amplify it. DeepSeek couldn't find anything but a 2x reduction, but reading the report of how it was done - https://github.com/KillaBoi/StealMyBitcoin-PoC - I was som", "context": "complete amateur", "use_case": "building exploit proof-of-concept", "switched_from": "", "story": "Chained PHP2Shell exploit roots 25 servers, including US Government servers"}, {"id": 804440898, "item_id": 50277556, "url": "https://news.ycombinator.com/item?id=50277556", "by": "cedws", "date": "2026-10-28", "stance": "neutral", "quote": "I use DeepSeek too but it\u2019s more of a workhorse model, requires a lot more guiding", "excerpt": "I\u2019m itching for more smart cheap models. I have a mid level Claude and ChatGPT subscription and I\u2019m blasting through my usage, and I\u2019m too cheap to pay for the higher tier. I use DeepSeek too but it\u2019s more of a workhorse model, requires a lot more guiding and sometimes struggles to get where I want it.", "context": "", "use_case": "cheap workhorse model", "switched_from": "", "story": "Gemini 4 Carbon"}, {"id": 804212720, "item_id": 50263295, "url": "https://news.ycombinator.com/item?id=50263295", "by": "ecommerceguy", "date": "2026-10-27", "stance": "recommend", "quote": "I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek", "excerpt": "I pay $10 a month for an AI Studio with an uncensored or de-censored Deepseek, and, quite a few others actually. Fun to use and cheap. I have no idea how anyone is going to profit off this stuff. I mean I make videos, images, analysis, presentations etc etc with this studio. $10 a month. Web interface. In addition to the 15 others I use randomly for generic searches.", "context": "", "use_case": "videos, images, analysis, presentations", "switched_from": "", "story": "There are now 5 inference providers for abliterated DeepSeek"}, {"id": 804082640, "item_id": 50255165, "url": "https://news.ycombinator.com/item?id=50255165", "by": "ddxv", "date": "2026-10-27", "stance": "neutral", "quote": "In regular work no, and I use it a lot.", "excerpt": "In regular work no, and I use it a lot. In purposefully asking it political questions yes. edit: Thinking about that now, that was over the webchat interface, so it might not have been related to the model, though I've seen various other related articles: https://www.crowdstrike.com/en-us/blog/crowdstrike-researchers-identify-hidden-vulnerabilities-ai-coded-software/", "context": "", "use_case": "general work", "switched_from": "", "story": "There are now 5 inference providers for abliterated DeepSeek"}, {"id": 804018448, "item_id": 50251153, "url": "https://news.ycombinator.com/item?id=50251153", "by": "dudisubekti", "date": "2026-10-26", "stance": "recommend", "quote": "It's dirt cheap, blazing fast, and is more than enough for most coding work.", "excerpt": "I use it at work. It's dirt cheap, blazing fast, and is more than enough for most coding work. If you live in a country with a weak currency, DeepSeek models are definitely a godsend.", "context": "", "use_case": "coding at work", "switched_from": "", "story": "GPT-6.1 Terra"}, {"id": 803950912, "item_id": 50246932, "url": "https://news.ycombinator.com/item?id=50246932", "by": "codebolt", "date": "2026-10-26", "stance": "recommend", "quote": "struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable", "excerpt": "I was playing around with OpenCode on Omarchy over the weekend, working on my Android app, and was struggling to spend a single dollar with DeepSeek. The quality difference from the frontier models I use at work was hardly noticeable (where I typically rack up at least a few hundred dollars per month).", "context": "works on an Android app", "use_case": "coding Android app", "switched_from": "", "story": "Anthropic S-1"}, {"id": 803918720, "item_id": 50244920, "url": "https://news.ycombinator.com/item?id=50244920", "by": "revolvingthrow", "date": "2026-10-26", "stance": "recommend", "quote": "Ultra-cheap models like Deepseek are also very nice if you need to burn billions and billions of tokens.", "excerpt": "I spend extra on a smartphone because I use the thing all the time, see it all the time, hold it in my hand all the time, it's my camera, music player, note taker and so on. For AI? Prompt comes in, tokens come out. The value of the 'ecosystem' is roughly zero to me. Switching costs between models are essentially zero. The UI and UX is the same whichever LLM runs underneath. If you do use llm some, then the $20 tier ", "context": "", "use_case": "high-volume token usage", "switched_from": "", "story": "Anthropic S-1"}, {"id": 803899730, "item_id": 50243733, "url": "https://news.ycombinator.com/item?id=50243733", "by": "Sphax", "date": "2026-10-26", "stance": "recommend", "quote": "These chinese models are simply good enough for the vast majority of tasks I run", "excerpt": "I use Opus 5.5 daily at work and a mix of GLM 5.3 (Flash or not), Deepseek v4 (flash or pro) at home and while Opus 5.5 is more capable, in my opinion it matters for less than 5% of the work i do. These chinese models are simply good enough for the vast majority of tasks I run and when they're not I use Astra or Opus 5.5. Yes, leaving Opus 5.5 on all the time is comfortable when you pay the 90\u20ac plan but it's just not", "context": "", "use_case": "home use", "switched_from": "", "story": "Anthropic S-1"}, {"id": 803831106, "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": 803443954, "item_id": 50215247, "url": "https://news.ycombinator.com/item?id=50215247", "by": "ricardobeat", "date": "2026-10-23", "stance": "neutral", "quote": "then handed off to DeepSeek or Minimax", "excerpt": "I had the lowest plan ($20/month I think) and it literally gave me fifteen, maybe twenty minutes tops, using the cheaper K3-256K variant. Basically enough to fill the context about 1.2x and that was it. It's a really smart model but way too expensive. The same $20 goes 100x further with Claude. I used it for the initial task planning while the plan didn't expire, then handed off to DeepSeek or Minimax.", "context": "", "use_case": "implementation after planning", "switched_from": "", "story": "Kimi K3.1 Pro and K3.1 Pro Highspeed"}, {"id": 802861571, "item_id": 50178848, "url": "https://news.ycombinator.com/item?id=50178848", "by": "_pdp_", "date": "2026-10-21", "stance": "neutral", "quote": "plenty of companies in UK and Europe that provide serverless access to GLM, DeepSeek etc.", "excerpt": "Someone asked me this morning if I consider Muse a privacy risk. Lol. I do. But to be honest this is all temporary. Muse, Grok, Dots will have users for the foreseeable future but there will be viable options for self-hosting too. I am already doing it https://github.com/pdparchitect/noodle/ The Hub powers the entire family inference needs through a single device - 4 users - 8 device about 50 agents - mixture of loca", "context": "", "use_case": "privacy-friendly cloud inference", "switched_from": "", "story": "Muse exposes users' personal and business secrets when they share conversations"}, {"id": 802860113, "item_id": 50178757, "url": "https://news.ycombinator.com/item?id=50178757", "by": "CharlieDigital", "date": "2026-10-21", "stance": "neutral", "quote": "Personal accounts I run the $17 Anthropic plan + $100 Codex Pro + DeepSeek.", "excerpt": "Yes, have done cross plat plan review. But why it works so well with Paseo is that it's automatic and multi-turn (agents kind of \"discuss\" (I say \"kind of\" because agents don't really discuss like you or I would with short, multi-turn speak-listen-pivot-repeat). Work tokens are severely limited to the extent that if I used Sonnet 5.5 for implementation, I'd be out 2h into a session (Team 5x). +$400/mo. at API rates o", "context": "", "use_case": "personal agentic dev", "switched_from": "", "story": "SlopCodeBench 10/2026 \u2013 Opus 5.5, Astra wins overall, xhigh considered harmful"}, {"id": 802796800, "item_id": 50174800, "url": "https://news.ycombinator.com/item?id=50174800", "by": "criley2", "date": "2026-10-21", "stance": "warn", "quote": "No need to spend twenty minutes pumping a few million tokens through deepseek regardless of how cheap it is to do so.", "excerpt": "No \"may be\" about it. There is no benefit to having bad models write your code, and there's no benefit to \"plan\" mode. Plan mode is an outdated 2025 concept that should be removed from standard harnesses. There are 2 scenarios for dumb builder agents: 1. Your \"planner\" actually implements the project almost completely, finds all the gotchas, solves all of the emergent problems, and fully plans the solution. 2. Your \"", "context": "", "use_case": "builder agent coding", "switched_from": "", "story": "SlopCodeBench 10/2026 \u2013 Opus 5.5, Astra wins overall, xhigh considered harmful"}, {"id": 802774881, "item_id": 50173430, "url": "https://news.ycombinator.com/item?id=50173430", "by": "apatheticonion", "date": "2026-10-20", "stance": "compare", "quote": "the running costs alone are more expensive than DeepSeek's API and the inference quality is much lower", "excerpt": "Here's an anecdote. I have repurposed my 6 year old 32gb M1 MacBook Pro to run Qwen 3.6 35b a3b and I used to give it nightly maintenance tasks that span my OSS github repos. I gave it mundane tasks like updating dependencies, writing changelogs, etc. It ran about 2 hours a night. My MBP consumes 0.1w idle and about 10w running inference. I pay $0.33/kwh. If it runs for 2 hours a day over 365 days - it will cost $2.4", "context": "", "use_case": "API inference", "switched_from": "", "story": "DeepInfer: High-performance inference framework for DeepSeek models"}, {"id": 802704705, "item_id": 50169044, "url": "https://news.ycombinator.com/item?id=50169044", "by": "solomonb", "date": "2026-10-20", "stance": "recommend", "quote": "I've been using it with deepseek for my homelab. Its pretty helpful.", "excerpt": "I made a project for this earlier this year: https://github.com/solomon-b/friendly-ghost/ You can can read from Loki or Journald and use regex or alert levels to filter what you pass to the LLM. I've been using it with deepseek for my homelab. Its pretty helpful.", "context": "homelab user", "use_case": "log analysis in homelab", "switched_from": "", "story": "Agentic Monitoring"}, {"id": 802609824, "item_id": 50163114, "url": "https://news.ycombinator.com/item?id=50163114", "by": "gigatexal", "date": "2026-10-20", "stance": "recommend", "quote": "Deepseek models are awesome. I use them all the time and they keep getting better.", "excerpt": "When you can\u2019t compete on price you ask your pedo Uncle Sam to help your businesses \u2026 Deepseek models are awesome. I use them all the time and they keep getting better.", "context": "", "use_case": "", "switched_from": "", "story": "DeepSeek banned from U.S. federal civilian agencies under new DHS order"}, {"id": 802560321, "item_id": 50160020, "url": "https://news.ycombinator.com/item?id=50160020", "by": "mlcruz", "date": "2026-10-20", "stance": "recommend", "quote": "Deepseek can do 90% of the job", "excerpt": "I think we keep getting used to intelligence improvements and the goal posts are always moving forward. I remember that close to an year ago i felt the same way about opus 4.5. And even if you can easily run ds or qwen locally to get very similar results to that, we moved to 5.5 being the new target. Deepseek can do 90% of the job but a claude reviwer/spec writer is a brutal improvement.", "context": "", "use_case": "coding", "switched_from": "", "story": "MiniMax M3.1 Flash"}, {"id": 802523680, "item_id": 50157730, "url": "https://news.ycombinator.com/item?id=50157730", "by": "surgical_fire", "date": "2026-10-19", "stance": "recommend", "quote": "DeepSeek probably could do everything I needed.", "excerpt": "> claude gives the edge I have no idea what you work with, but for every single programming job I had in more than 2 decades, from the lamest consultancy to FAANG, you don't need the Edge. DeepSeek probably could do everything I needed. Would I pay for LLMs? Sure, but not much more than 30-50 bucks range per month. I can affort 1k/m, but no way in hell I would accept it. I would rather just not use LLMs (it's not lik", "context": "20+ years programming", "use_case": "programming work", "switched_from": "", "story": "The obscure trust looming over Anthropic's IPO"}, {"id": 802515073, "item_id": 50157192, "url": "https://news.ycombinator.com/item?id=50157192", "by": "Rebelgecko", "date": "2026-10-19", "stance": "recommend", "quote": "I'll be sticking with Deepseek v4 for sure.", "excerpt": "I used Space Bunny from some providers that offered it for free and it was a very much a \"you get what you pay for\" experience. If the release is comparable, I'll be sticking with Deepseek v4 for sure.", "context": "", "use_case": "", "switched_from": "", "story": "MiniMax M3.1 Flash"}, {"id": 802256304, "item_id": 50141019, "url": "https://news.ycombinator.com/item?id=50141019", "by": "pimeys", "date": "2026-10-18", "stance": "recommend", "quote": "now I just do planning with DeepSeek too and don't see much difference", "excerpt": "I also used to plan with more expensive models like you did, but now I just do planning with DeepSeek too and don't see much difference. I use the omp plan mode and iterate on the text until I'm happy with it. Then clear context, give the plan to DeepSeek again for implementation. If it's mission-critical code, do a security review with GLM 5.3. Otherwise go manually through the code and ask for changes if needed. Wo", "context": "", "use_case": "planning and implementation", "switched_from": "more expensive models", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802166562, "item_id": 50135410, "url": "https://news.ycombinator.com/item?id=50135410", "by": "aizk", "date": "2026-10-17", "stance": "neutral", "quote": "for the really hard functions I don't use deepseek I use opus managing other opus agents", "excerpt": "Caveat: I get a budget of free ai tokens at work, as I work for an open source AI inference provider. My goal is to reverse engineer an entire gamecube game -- I'm pushing an existing community decomp that was stalled to the finish line, solo, with lots of AI help. This work was directly inspired by https://decomp.dev/ -- which has just shut down a few days ago because of, ironically... people like me who enjoy decom", "context": "works for an open source AI inference provider", "use_case": "decompilation agents", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802115936, "item_id": 50132246, "url": "https://news.ycombinator.com/item?id=50132246", "by": "sligbad", "date": "2026-10-17", "stance": "recommend", "quote": "daily driver since V4", "excerpt": "Once read, I think from G Huntley, \"pick a model and stick with it,\" they have been making that very easy, daily driver since V4.", "context": "", "use_case": "daily AI model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802114336, "item_id": 50132146, "url": "https://news.ycombinator.com/item?id=50132146", "by": "neya", "date": "2026-10-17", "stance": "recommend", "quote": "I love DS and am not looking back.", "excerpt": "I second this, I love DS and am not looking back. For Elixir, not a lot of models are good, but DS just blazes through it.", "context": "", "use_case": "Elixir coding", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802108304, "item_id": 50131769, "url": "https://news.ycombinator.com/item?id=50131769", "by": "drewnick", "date": "2026-10-17", "stance": "recommend", "quote": "Works very well and feels almost free.", "excerpt": "I use deepseek via fireworks using Claude code. When I type deepseek on my CLI it launches Claude with my fireworks API key. Works very well and feels almost free.", "context": "", "use_case": "CLI coding agent model", "switched_from": "", "story": "DeepSeek: DeepSeek v4.1 Flash 1017"}, {"id": 802107809, "item_id": 50131738, "url": "https://news.ycombinator.com/item?id=50131738", "by": "aizk", "date": "2026-10-17", "stance": "recommend", "quote": "I use Opus 5.5 and swarms of about 50 deepseek models to reverse engineer old videos games.", "excerpt": "I use Opus 5.5 and swarms of about 50 deepseek models to reverse engineer old videos games. The scale is roughly 600k loc generated from pure asm, simply by knowing what the old compiler is. Truly staggering stuff.", "context": "", "use_case": "reverse engineering old games", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802106800, "item_id": 50131675, "url": "https://news.ycombinator.com/item?id=50131675", "by": "dw_arthur", "date": "2026-10-17", "stance": "recommend", "quote": "I am a huge DeepSeek fan", "excerpt": "I am a huge DeepSeek fan, but it cannot come close to Opus 5.5 on UI design. A $20 Claude subscription that I use some months and cancel the others along with Deepseek API is great.", "context": "", "use_case": "coding via API", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802085440, "item_id": 50130340, "url": "https://news.ycombinator.com/item?id=50130340", "by": "davedx", "date": "2026-10-17", "stance": "recommend", "quote": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well.", "excerpt": "I have started phasing into DeepSeek v4 (Dwarfstar quant from Antirez) as my daily driver and it's going very well. It can one shot iOS apps on my Mac Studio, it's a little slow and sometimes needs retries but can definitely do medium difficulty real work for me. I didn't try the last flash model because I would need to use SSD streaming (I have 128 GB), but the benches for this release are so compelling (and it usin", "context": "runs local models on a Mac Studio with 128 GB", "use_case": "local coding model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802075264, "item_id": 50129704, "url": "https://news.ycombinator.com/item?id=50129704", "by": "ytxm", "date": "2026-10-17", "stance": "neutral", "quote": "I gave it a kick and got mixed results.", "excerpt": "I gave it a kick and got mixed results. For example, searching for a Microsoft product couldn't find its landing page, and searching for a niche open source package landed in the Debian repos for the top hits, but the home page was in the first 10 hits. Lovely to see they're doing their own search though, I love how many folk are actively working on search clones at the moment.", "context": "", "use_case": "web search API", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802069840, "item_id": 50129365, "url": "https://news.ycombinator.com/item?id=50129365", "by": "ytxm", "date": "2026-10-17", "stance": "recommend", "quote": "I use it for literally everything.", "excerpt": "It passed my threshold for not caring to try new models back with DS4, I use it for literally everything. Announcements of new models in the series always feel curious, because it doesn't really feel like there's much left to fix that would improve on what's already there for any task I know of. It's a bit like obsessing over patch releases of some software after a major upgrade already solved the original problem lo", "context": "", "use_case": "general-purpose AI model", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 802068865, "item_id": 50129304, "url": "https://news.ycombinator.com/item?id=50129304", "by": "egeozcan", "date": "2026-10-17", "stance": "recommend", "quote": "the rest can be happily done with deepseek or glm flash", "excerpt": "I use Opus for planning, through API on medium thinking, it doesn't cost anything significant and the rest can be happily done with deepseek or glm flash. Using sol 6.1 for review loops at the end. All running on pi. A very complex task tends to cost like 10$ max. It's comparable to a subscription with enormous flexibility. It's a steal.", "context": "", "use_case": "implementation", "switched_from": "", "story": "DeepSeek-v4.1-Flash-1017"}, {"id": 801780064, "item_id": 50111254, "url": "https://news.ycombinator.com/item?id=50111254", "by": "Shebanator", "date": "2026-10-15", "stance": "recommend", "quote": "I switched to deepseek because it was giving me better coding results than GLM, and for far less cost.", "excerpt": "Model sizes aren't directly comparable, and they aren't even that relevant. I care about result quality, cost, and speed, not size. I switched to deepseek because it was giving me better coding results than GLM, and for far less cost. GLM is still better at planning.", "context": "", "use_case": "coding", "switched_from": "GLM", "story": "Muse Spark Open"}, {"id": 800529312, "item_id": 50033082, "url": "https://news.ycombinator.com/item?id=50033082", "by": "pmkary", "date": "2026-10-10", "stance": "recommend", "quote": "somehow for me only deepseek works", "excerpt": "You're reading it in an unintended way. What I meant was \"I used all of the models multiple times on exact same prompt and somehow for me only deepseek works\". I haven't had my hands on Astra or Opus 5.5, they perhaps will have good results, but the others I tried at least once a month and horrible results. So I'm stuck with DeepSeek and don't use the other things.", "context": "", "use_case": "", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 800365041, "item_id": 50022815, "url": "https://news.ycombinator.com/item?id=50022815", "by": "pimeys", "date": "2026-10-09", "stance": "compare", "quote": "DeepSeek is great if you have a plan and you let it go.", "excerpt": "I like to do some discussions and back and forth with GLM. It doesn't hallucinate as much or gaslight like DeepSeek does. DeepSeek is great if you have a plan and you let it go. Also can plan a bit. But it has this sparse attention and especially the index_topk is too small in common deployments so it doesn't \"see\" all the context. And it then imagines things which is annoying if you chat with it about stuff you deve", "context": "", "use_case": "executing a defined plan", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 800292913, "item_id": 50018307, "url": "https://news.ycombinator.com/item?id=50018307", "by": "tacomagick", "date": "2026-10-09", "stance": "neutral", "quote": "I use Deepseek myself.", "excerpt": "My organization provides free Gemini access, I use Deepseek myself. I have a Opencode Go subscription for the time being because it provides both ZDR with their agreement with Deepseek and 10$ of subscription turns to 60$ of credits for Deepseek as part of their release discount.", "context": "", "use_case": "personal AI model", "switched_from": "", "story": "Ask HN: Are you a subscribed LLM or a locally deployed open-source model?"}, {"id": 800239954, "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": 800214497, "item_id": 50013406, "url": "https://news.ycombinator.com/item?id=50013406", "by": "denkmoon", "date": "2026-10-08", "stance": "recommend", "quote": "I don't see any reason GLM and DeepSeek won't eventually be as good as Claude", "excerpt": "I use the frontier openai/anthropic models at work but exclusively open weight models (on cloud/hosted inference) for personal stuff and I think about it like this; 1) I don't see any reason GLM and DeepSeek won't eventually be as good as Claude, it's just a matter of time and 2) the open models are well and truly capable enough for most of what I'd want to do. I don't need nor want an LLM chewing away on a horrible ", "context": "uses frontier models at work", "use_case": "personal projects", "switched_from": "", "story": "Why isn't the industry freaking out about DeepSeek 4.1 Flash?"}, {"id": 800170944, "item_id": 50010684, "url": "https://news.ycombinator.com/item?id=50010684", "by": "celrod", "date": "2026-10-08", "stance": "recommend", "quote": "I liked DeepSeek and GLM when I used them.", "excerpt": "Using Artificial Analysis Model | Reasoning | Intelligence Index | Artificial Analysis million output tokens for the intelligence index -|-|-|- Step 5 | ? | 44 | 160 GLM 5.3 | Max | 45 | 210 MiMo V2.6 Pro | ? | 46 | 140 Kimi K3 | Max | 44 | 160 Qwen Max 0902 | ? | 45 | 190 DeepSeek 4.1 Flash | Max | 39 | 250 GLM 5.3-flash | Max | 42 | 180 GPT-6 Astra | Low | 46 | 10 It looks reasonable by open model standards. This d", "context": "", "use_case": "", "switched_from": "", "story": "Step 5 Preview, a 1M-context MoE from StepFun, shows up on OpenRouter"}, {"id": 800014817, "item_id": 50000926, "url": "https://news.ycombinator.com/item?id=50000926", "by": "zyk_computer", "date": "2026-10-08", "stance": "recommend", "quote": "I used to always use glm or deepseek instead of Haiku", "excerpt": "I used to always use glm or deepseek instead of Haiku, but now I want to give it a try.", "context": "", "use_case": "", "switched_from": "", "story": "Claude Haiku 5.5"}, {"id": 799953682, "item_id": 49997105, "url": "https://news.ycombinator.com/item?id=49997105", "by": "wyrdcurt", "date": "2026-10-07", "stance": "neutral", "quote": "along with some MiMo and DeepSeek", "excerpt": "Some people/organizations are ideologically opposed to using Chinese models. Not me, I use GLM-5.3-Flash for almost everything (the subscription-subsidized pricing on a legacy Z.ai plan makes it the best value model by a wide margin), along with some MiMo and DeepSeek. Still, I use Luna for certain tasks where speed is more valuable than performance; I can see this new Haiku displacing Luna for those. If you mean Hai", "context": "", "use_case": "", "switched_from": "", "story": "Claude Haiku 5.5"}, {"id": 798373696, "item_id": 49898356, "url": "https://news.ycombinator.com/item?id=49898356", "by": "proxysna", "date": "2026-09-29", "stance": "recommend", "quote": "Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper", "excerpt": "I am yet to spend $200 on deepseek this year. Not sure what kind of usage can justify $200/month of either openai or anthropic, i'm not even talking about $500. Deepseek is faster, IMO intelligence difference is negligible and it so much cheaper that i no longer care about how much i use it. I never hit any daily/weekly quota or anything like that while working or tinkering. At this point i am OK with being 6 months ", "context": "", "use_case": "everyday coding and tinkering", "switched_from": "", "story": "GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price"}, {"id": 797168864, "item_id": 49823054, "url": "https://news.ycombinator.com/item?id=49823054", "by": "switchbak", "date": "2026-09-23", "stance": "recommend", "quote": "I use Deepseek for some home stuff, and it's as you describe - fast, consistent.", "excerpt": "This is so true, I use frontier models at work and the variance in the day to day experience can be dramatic. I use Deepseek for some home stuff, and it's as you describe - fast, consistent. Not brainiac level, but when you drive it well it just tends to execute reliably. I have seen it fail consistently on very hard stuff, but I'm certainly ok with having to special case occasionally. Especially if those occasions k", "context": "uses frontier models at work", "use_case": "home projects", "switched_from": "", "story": "Claude Opus 5.5"}, {"id": 796993488, "item_id": 49812093, "url": "https://news.ycombinator.com/item?id=49812093", "by": "knapcio", "date": "2026-09-23", "stance": "recommend", "quote": "Then I switched to DeepSeek and haven\u2019t noticed any drop in quality. It\u2019s fast at around 300 TPS and cheap.", "excerpt": "I\u2019ve been working with frontier models for the last year on a large real-world project with multiple apps. Eventually they started becoming unreliable. Simple UI issues, usage limits and overengineering became constant problems. I had to come up with a robust process: task analysis, user intent analysis, implementation and testing. All of these steps loop when needed with 15-20 steps per task in total. This finally g", "context": "works on a large real-world multi-app project", "use_case": "coding on large project", "switched_from": "", "story": "Claude Opus 5.5"}, {"id": 796953681, "item_id": 49809605, "url": "https://news.ycombinator.com/item?id=49809605", "by": "abalashov", "date": "2026-09-22", "stance": "recommend", "quote": "very cheap large-scale code analyses and things that require mega context windows", "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": "large-scale code analysis", "switched_from": "", "story": "MiMo-v2.6-Pro: Intelligence, Performance and Price Analysis"}, {"id": 796680674, "item_id": 49792542, "url": "https://news.ycombinator.com/item?id=49792542", "by": "RussianCow", "date": "2026-09-21", "stance": "warn", "quote": "The downside is they train their models on your input, which might be a deal-breaker for many users (as it is for me).", "excerpt": "For caching, only if you don't specify your preferred providers and let OpenRouter route each request itself. I have stuff like this in my OpenCode config for each model I use and I regularly get ~90-95% cache hit rates. \"order\": [\"relace\", \"coreweave\", \"novita\", \"baseten\", \"together\"], \"allow_fallbacks\": false It still won't be quite as high as you'd get by just using DeepSeek because occasionally a request will fai", "context": "", "use_case": "direct API use", "switched_from": "", "story": "Grok 4.7"}]}