{"slug": "gemini-pro", "name": "Gemini Pro", "category": "ai", "category_name": "AI models and assistants", "description": "Google's flagship Gemini large language model, available via the Gemini app, AI Studio and API.", "recommenders": 31, "commenters": 67, "warners": 10, "score": 25.179, "rising": -0.85, "first": "2025-11-04", "latest": "2026-10-29", "why": "Liked for everyday chat and deep, obscure knowledge, and for cleaning up, refactoring and planning code alongside other models.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 30, "neutral": 27, "compare": 3, "warn": 10}, "spark": [3, 4, 5, 3, 2, 1, 2, 2, 1, 0, 5, 2], "listed": true, "loved": ["Knowledge and chat", "Refactoring and review"], "caveat": "Quality degradation", "short_quote": "I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff.", "icon": true, "domain": null, "github_repo": null, "official_url": "https://gemini.google.com", "pricing": "freemium", "aliases": ["Gemini 3.1 Pro", "Gemini 3 Pro", "Gemini Pro", "Gemini 2.5 Pro", "Gemini 3.0 Pro", "Gemini 3 Pro Preview", "Gemini-3-Pro"], "verdict": [{"text": "Liked for everyday chat and deep, obscure knowledge,", "cites": [{"id": 796659840, "item_id": 49791240, "url": "https://news.ycombinator.com/item?id=49791240", "by": "moojacob", "date": "2026-09-21", "stance": "recommend", "quote": "I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff.", "excerpt": "I'll have to try Gemini Flash for coding. The reason I haven't I used Gemini for coding is last time I tried it couldn't call tools very well. I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff. I'll double check something Gemini said and find the source is deep inside a hard to access scientific paper. Google just has the best index of the internet.", "context": "", "use_case": "chat and research", "switched_from": "", "story": "Grok 4.7"}, {"id": 769812528, "item_id": 48113283, "url": "https://news.ycombinator.com/item?id=48113283", "by": "xiphias2", "date": "2026-05-12", "stance": "recommend", "quote": "I love using the Gemini pro model for quick questions", "excerpt": "Google needs to beat OpenAI and Antropic in coding models because that's where the big money is going. I love using the Gemini pro model for quick questions, but that's not where I'm spending the real money. They have so many great software engineers but unable to use them to speed up coding AI research. Hopefully with Sergey's focus it will get better. This cursor thing is just another experiment nobody cares about.", "context": "", "use_case": "quick questions", "switched_from": "", "story": "AI Pointer from DeepMind"}, {"id": 794703424, "item_id": 49668964, "url": "https://news.ycombinator.com/item?id=49668964", "by": "grahamnorton39", "date": "2026-09-12", "stance": "recommend", "quote": "2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text", "excerpt": "Might be missing something \u2014- are there any issues with Gemini 3.1 Pro that aren\u2019t there in 2.5? I agree, though. 2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text (and videos, and images, and audio too iirc?). Basically unlimited access to it too via AI Studio. I used it for processing and transforming bucketloads of data, ingesting masses of transcripts and converting them to flash", "context": "", "use_case": "processing transcripts into flashcards", "switched_from": "", "story": "Ask HN: Did Google kill its enterprise workhorse model?"}]}, {"text": " and for cleaning up, refactoring and planning code alongside other models.", "cites": [{"id": 757589697, "item_id": 47349356, "url": "https://news.ycombinator.com/item?id=47349356", "by": "stared", "date": "2026-03-12", "stance": "recommend", "quote": "My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up.", "excerpt": "I spend much more time refactoring that creating features (though, it is getting better with each model). My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up. I feel that doing it just one-stage is rarely enough. A lot of prompts about finding the right level of abstraction, DRY, etc. An earlier example (Opus 4.5 + Gemini 3 Pro) is here: https://github.com/stared/sc2-balanc", "context": "", "use_case": "refactoring and cleanup", "switched_from": "", "story": "Many SWE-bench-Passing PRs would not be merged"}, {"id": 783896464, "item_id": 48993529, "url": "https://news.ycombinator.com/item?id=48993529", "by": "glimshe", "date": "2026-07-21", "stance": "recommend", "quote": "I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review", "excerpt": "I respectfully disagree, at least for most tasks. 3.1 Pro: while it's coding performance is mediocre, a lot of coding work requires minimum actual thinking. I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review, language questions (\"is there a better way to write this code block?\"). I don't have a corporation behind me so costs matter. Considering that I'm a Pro subsc", "context": "Pro subscriber without a corporation behind them", "use_case": "light refactoring and boilerplate", "switched_from": "", "story": "Gemini 3.6 Flash"}, {"id": 753716576, "item_id": 47107286, "url": "https://news.ycombinator.com/item?id=47107286", "by": "brandall10", "date": "2026-02-21", "stance": "recommend", "quote": "I have a planner skill I use w/ Gemini Pro to generate new plans.", "excerpt": "I go a bit further than this and have had incredible success with 3 doc types: - Specs: these are generally static, but updatable as the project evolves. And they're broken out to an index file that gives a project overview, a high-level arch file, and files for all the main modules. Roughly ~1k lines of spec for 10k lines of code - Plans: these are the output of a planning session with an LLM. They point to the asso", "context": "", "use_case": "generating plans", "switched_from": "", "story": "How I use Claude Code: Separation of planning and execution"}]}, {"text": " The most common complaints are that quality seems to drop after launch", "cites": [{"id": 804462736, "item_id": 50278921, "url": "https://news.ycombinator.com/item?id=50278921", "by": "5555watch", "date": "2026-10-28", "stance": "recommend", "quote": "I loved Gemini 3.1 Pro", "excerpt": "I loved Gemini 3.1 Pro and it was very sad when they nerfed it at around May this year. I think it did pretty good as a chat, and quite decent on Antigravity (tried very little). Been sitting on 5.6 / Astra and Codex since. Occasionally do bring up the 3.1 Pro through the API, because I love the writing style, but keep noticing that I'm much more productive with the frontier, despite the love to Gemini. Tried 4 Argon", "context": "", "use_case": "chat and writing", "switched_from": "", "story": "Gemini 4 Carbon"}, {"id": 753690310, "item_id": 47105644, "url": "https://news.ycombinator.com/item?id=47105644", "by": "dudeinhawaii", "date": "2026-02-21", "stance": "warn", "quote": "Gemini 3 Pro Preview went from exceptional in the first month to mediocre", "excerpt": "After 2 days of giving it a go, I find that Gemini CLI is still considerably worse than both Codex and Claude Code. The model itself also has strange behaviors that seem like it gets randomly replaced with Gemini-3-Flash or something else. I'll explain. Once agentic coding was a bust, I gave it a run as a daily driver for AI assistant. It performed fairly well but then began behaving strangely. It would lose context ", "context": "", "use_case": "", "switched_from": "", "story": "Gemini 3.1 Pro Preview"}, {"id": 754945681, "item_id": 47184105, "url": "https://news.ycombinator.com/item?id=47184105", "by": "idoxer", "date": "2026-02-27", "stance": "warn", "quote": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released", "excerpt": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released, now the Gemini 3 Flash is having peak times often. Also cause google to continuously reducing free tiers to stable their infrastructure And it's also seems to be worse from here, probably for all major providers", "context": "", "use_case": "", "switched_from": "", "story": "Gemini's 10 days of different outages and increasing high demand"}]}, {"text": " and that tool calling in agentic coding is unreliable.", "cites": [{"id": 796666673, "item_id": 49791667, "url": "https://news.ycombinator.com/item?id=49791667", "by": "haellsigh", "date": "2026-09-21", "stance": "warn", "quote": "the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls", "excerpt": "I remember the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls. They really redeemed themselves with the 3.5 Flash series, it's great. I love how fast it is. Now if only they could offer a subscription service for it...", "context": "", "use_case": "text", "switched_from": "", "story": "Grok 4.7"}, {"id": 737029073, "item_id": 46064317, "url": "https://news.ycombinator.com/item?id=46064317", "by": "terrywang", "date": "2025-11-26", "stance": "warn", "quote": "relatively weak tool calling capability of the Gemini 2.5 Pro model", "excerpt": "Gemini CLI at this stage isn't good at complex coding tasks (vs. Claude Code, Codex, Cursor CLI, Qoder CLI, etc.). Mostly because of the simple ReAct loop, compounded by relatively weak tool calling capability of the Gemini 2.5 Pro model. > I haven't tried complex coding tasks using Gemini 3.0 Pro Preview yet. I reckon it won't be materially different. Gemini CLI is open source and being actively developed, which is ", "context": "", "use_case": "coding", "switched_from": "", "story": "Gemini CLI tips and tricks for agentic coding"}, {"id": 744602704, "item_id": 46537669, "url": "https://news.ycombinator.com/item?id=46537669", "by": "theturtletalks", "date": "2026-01-07", "stance": "warn", "quote": "Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem", "excerpt": "All of this seems like manufactured hype for Gemini. I use GPT-5.2, Opus 4.5, and Gemini 3 flash and pro with Droid CLI and Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem, and still fails to call tools consistently (sometimes the whole thread is corrupted and you can\u2019t rewind and use another model). Terminal Bench supports my findings, GPT-5.2 and", "context": "", "use_case": "agentic coding", "switched_from": "", "story": "Google Got Its Groove Back and Edged Ahead of OpenAI"}]}], "praise": [{"label": "Knowledge and chat", "summary": "Commenters use it for everyday questions and chat and say it knows a lot, including obscure facts.", "quotes": [{"id": 796659840, "item_id": 49791240, "url": "https://news.ycombinator.com/item?id=49791240", "by": "moojacob", "date": "2026-09-21", "stance": "recommend", "quote": "I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff.", "excerpt": "I'll have to try Gemini Flash for coding. The reason I haven't I used Gemini for coding is last time I tried it couldn't call tools very well. I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff. I'll double check something Gemini said and find the source is deep inside a hard to access scientific paper. Google just has the best index of the internet.", "context": "", "use_case": "chat and research", "switched_from": "", "story": "Grok 4.7"}, {"id": 769812528, "item_id": 48113283, "url": "https://news.ycombinator.com/item?id=48113283", "by": "xiphias2", "date": "2026-05-12", "stance": "recommend", "quote": "I love using the Gemini pro model for quick questions", "excerpt": "Google needs to beat OpenAI and Antropic in coding models because that's where the big money is going. I love using the Gemini pro model for quick questions, but that's not where I'm spending the real money. They have so many great software engineers but unable to use them to speed up coding AI research. Hopefully with Sergey's focus it will get better. This cursor thing is just another experiment nobody cares about.", "context": "", "use_case": "quick questions", "switched_from": "", "story": "AI Pointer from DeepMind"}, {"id": 794703424, "item_id": 49668964, "url": "https://news.ycombinator.com/item?id=49668964", "by": "grahamnorton39", "date": "2026-09-12", "stance": "recommend", "quote": "2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text", "excerpt": "Might be missing something \u2014- are there any issues with Gemini 3.1 Pro that aren\u2019t there in 2.5? I agree, though. 2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text (and videos, and images, and audio too iirc?). Basically unlimited access to it too via AI Studio. I used it for processing and transforming bucketloads of data, ingesting masses of transcripts and converting them to flash", "context": "", "use_case": "processing transcripts into flashcards", "switched_from": "", "story": "Ask HN: Did Google kill its enterprise workhorse model?"}]}, {"label": "Refactoring and review", "summary": "Used for cleanup, light refactoring, boilerplate and code review, often after another model writes the code.", "quotes": [{"id": 757589697, "item_id": 47349356, "url": "https://news.ycombinator.com/item?id=47349356", "by": "stared", "date": "2026-03-12", "stance": "recommend", "quote": "My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up.", "excerpt": "I spend much more time refactoring that creating features (though, it is getting better with each model). My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up. I feel that doing it just one-stage is rarely enough. A lot of prompts about finding the right level of abstraction, DRY, etc. An earlier example (Opus 4.5 + Gemini 3 Pro) is here: https://github.com/stared/sc2-balanc", "context": "", "use_case": "refactoring and cleanup", "switched_from": "", "story": "Many SWE-bench-Passing PRs would not be merged"}, {"id": 783896464, "item_id": 48993529, "url": "https://news.ycombinator.com/item?id=48993529", "by": "glimshe", "date": "2026-07-21", "stance": "recommend", "quote": "I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review", "excerpt": "I respectfully disagree, at least for most tasks. 3.1 Pro: while it's coding performance is mediocre, a lot of coding work requires minimum actual thinking. I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review, language questions (\"is there a better way to write this code block?\"). I don't have a corporation behind me so costs matter. Considering that I'm a Pro subsc", "context": "Pro subscriber without a corporation behind them", "use_case": "light refactoring and boilerplate", "switched_from": "", "story": "Gemini 3.6 Flash"}, {"id": 746079744, "item_id": 46629984, "url": "https://news.ycombinator.com/item?id=46629984", "by": "esperent", "date": "2026-01-15", "stance": "recommend", "quote": "3 pro is very good - a bit behind codex but still very useful.", "excerpt": "I very much doubt this. I've been using gemini whenever I get hit by codex limits over the past week. 3 pro is very good - a bit behind codex but still very useful. I've tried 3 flash several times and each time what I got back was complete garbage. After the third or fourth attempt I stopped trying.", "context": "", "use_case": "coding when Codex limits hit", "switched_from": "", "story": "Ask HN: What is best way to provide continuous context to models?"}]}, {"label": "Reasoning and accuracy", "summary": "Some call it the best model they have tried for troubleshooting logic, reverse engineering and medical questions.", "quotes": [{"id": 798662544, "item_id": 49916409, "url": "https://news.ycombinator.com/item?id=49916409", "by": "landdate", "date": "2026-10-01", "stance": "recommend", "quote": "Its easily the best model I have tried. And yes I am talking about 3.1 pro.", "excerpt": "I don't like AI, but it's very enjoyable for me to see my predictions on the success of gemini come to fruition. I only use gemini, and while I don't use it for actually writing up code, I use it to help me troibleshoot my logic and help find bugs. Its easily the best model I have tried. And yes I am talking about 3.1 pro. Also I have found gemini the only model to be the least likely to douse me in flattery, and wil", "context": "", "use_case": "troubleshooting logic and finding bugs", "switched_from": "", "story": "Gemini 4 Argon"}, {"id": 741075776, "item_id": 46317236, "url": "https://news.ycombinator.com/item?id=46317236", "by": "lkt", "date": "2025-12-18", "stance": "recommend", "quote": "found Gemini 3 Pro to be the best model for that, surprisingly much better than Opus 4.5", "excerpt": "I've been doing some reverse engineering recently and have found Gemini 3 Pro to be the best model for that, surprisingly much better than Opus 4.5. Maybe it's time to give Codex a try", "context": "", "use_case": "reverse engineering", "switched_from": "", "story": "GPT-5.2-Codex"}, {"id": 794108289, "item_id": 49631768, "url": "https://news.ycombinator.com/item?id=49631768", "by": "toshvelaga", "date": "2026-09-09", "stance": "recommend", "quote": "Gemini 3.8 flash and gemini 3.1 pro seem to lead across the other LLMs.", "excerpt": "There are no sites like artificialanalysis.ai or arena.ai for healthcare benchmarks so I decided to build it. I ran the leading models from openai, anthropic, and google on MedXpertQA: https://github.com/TsinghuaC3I/MedXpertQA#leaderboard The results show scores across organ systems and across diagnosis, treatment, and basic science. Gemini 3.8 flash and gemini 3.1 pro seem to lead across the other LLMs. The last tim", "context": "built a healthcare LLM benchmark", "use_case": "medical question benchmark", "switched_from": "", "story": "Benchmarked how the latest LLMs perform on expert level medical questions"}]}, {"label": "Planning", "summary": "Used for planning, problem solving and generating project plans.", "quotes": [{"id": 757094946, "item_id": 47318434, "url": "https://news.ycombinator.com/item?id=47318434", "by": "radio879", "date": "2026-03-09", "stance": "recommend", "quote": "use pro for planning, fixing, problem solving, all the difficult stuff", "excerpt": "With all of the talk about \"LLMs are non deterministic\" I am trying to build my own tool (at this exact moment, since I seem to have gotten really good at this) which does some of what these OpenClaw/Clawdbots do for people, but instead of having these LLMs/agents handle everything I am trying to do as much of it as pure deterministic code as possible. I test things once, like daily automations, and they will repeat ", "context": "works with LLMs daily for 8-15 hours", "use_case": "planning and problem solving", "switched_from": "", "story": "Ask HN: Read\u2011only LLM tool for email triage and knowledge extraction?"}, {"id": 753716576, "item_id": 47107286, "url": "https://news.ycombinator.com/item?id=47107286", "by": "brandall10", "date": "2026-02-21", "stance": "recommend", "quote": "I have a planner skill I use w/ Gemini Pro to generate new plans.", "excerpt": "I go a bit further than this and have had incredible success with 3 doc types: - Specs: these are generally static, but updatable as the project evolves. And they're broken out to an index file that gives a project overview, a high-level arch file, and files for all the main modules. Roughly ~1k lines of spec for 10k lines of code - Plans: these are the output of a planning session with an LLM. They point to the asso", "context": "", "use_case": "generating plans", "switched_from": "", "story": "How I use Claude Code: Separation of planning and execution"}]}, {"label": "Long context", "summary": "Its large context window lets it take in big codebases and large volumes of text.", "quotes": [{"id": 747507232, "item_id": 46719202, "url": "https://news.ycombinator.com/item?id=46719202", "by": "erichocean", "date": "2026-01-22", "stance": "recommend", "quote": "I require Gemini's long context size because my approach leans heavily on in context learning to produce correct code.", "excerpt": "I've had AI write ~100% of my code for the last 7 months, but I acted as the \"agent\" so the AI had very high levels of direction, and I approved all code changes at every step, including during debugging. Mostly Gemini Pro 2.5 (and now Gemini Pro 3) and mostly Clojure and/or Java, with some JavaScript/Python. I require Gemini's long context size because my approach leans heavily on in context learning to produce corr", "context": "has AI write ~100% of code, Clojure/Java", "use_case": "writing code with human-in-the-loop", "switched_from": "", "story": "Ask HN: Do you have any evidence that agentic coding works?"}, {"id": 794703424, "item_id": 49668964, "url": "https://news.ycombinator.com/item?id=49668964", "by": "grahamnorton39", "date": "2026-09-12", "stance": "recommend", "quote": "2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text", "excerpt": "Might be missing something \u2014- are there any issues with Gemini 3.1 Pro that aren\u2019t there in 2.5? I agree, though. 2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text (and videos, and images, and audio too iirc?). Basically unlimited access to it too via AI Studio. I used it for processing and transforming bucketloads of data, ingesting masses of transcripts and converting them to flash", "context": "", "use_case": "processing transcripts into flashcards", "switched_from": "", "story": "Ask HN: Did Google kill its enterprise workhorse model?"}]}, {"label": "UI design", "summary": "Commenters pick it for UI polish and say its design work is above other models.", "quotes": [{"id": 763364706, "item_id": 47710294, "url": "https://news.ycombinator.com/item?id=47710294", "by": "dokdev", "date": "2026-04-09", "stance": "recommend", "quote": "If I need some UI polish, I switch to gemini 3.1 pro.", "excerpt": "Cursor $20 subscription is also very good. auto is usually decent. I use composer 2 for heavy tasks and it does good job. If I need some UI polish, I switch to gemini 3.1 pro. The trick is to stay at auto/composer 2. But if you try to use other premium models too much, then you\u2019ll hit the limit as well.", "context": "", "use_case": "UI polish", "switched_from": "", "story": "Claude Code limits are starting to feel like a psychological trick"}, {"id": 780134930, "item_id": 48758433, "url": "https://news.ycombinator.com/item?id=48758433", "by": "muzani", "date": "2026-07-02", "stance": "recommend", "quote": "Gemini 3 Pro is actually at a mid designer level.", "excerpt": "In the last 12 months? Antigravity started the workflow where you give a list of things to do and it goes off and does those things without supervision, including drafting and testing edge cases. It can even spin up images. Fable is the latest form of this workflow. Gemini 3 Pro is actually at a mid designer level. None of the others are even at a junior human design level. Sonnet 4.5 was, for a brief moment, creativ", "context": "", "use_case": "design", "switched_from": "", "story": "Ask HN: Does anybody still FEEL improvements between latest LLMs for coding?"}]}], "caveats": [{"label": "Quality degradation", "summary": "Several users say it got worse after launch or behaved as if it had been swapped for a weaker model.", "quotes": [{"id": 804462736, "item_id": 50278921, "url": "https://news.ycombinator.com/item?id=50278921", "by": "5555watch", "date": "2026-10-28", "stance": "recommend", "quote": "I loved Gemini 3.1 Pro", "excerpt": "I loved Gemini 3.1 Pro and it was very sad when they nerfed it at around May this year. I think it did pretty good as a chat, and quite decent on Antigravity (tried very little). Been sitting on 5.6 / Astra and Codex since. Occasionally do bring up the 3.1 Pro through the API, because I love the writing style, but keep noticing that I'm much more productive with the frontier, despite the love to Gemini. Tried 4 Argon", "context": "", "use_case": "chat and writing", "switched_from": "", "story": "Gemini 4 Carbon"}, {"id": 753690310, "item_id": 47105644, "url": "https://news.ycombinator.com/item?id=47105644", "by": "dudeinhawaii", "date": "2026-02-21", "stance": "warn", "quote": "Gemini 3 Pro Preview went from exceptional in the first month to mediocre", "excerpt": "After 2 days of giving it a go, I find that Gemini CLI is still considerably worse than both Codex and Claude Code. The model itself also has strange behaviors that seem like it gets randomly replaced with Gemini-3-Flash or something else. I'll explain. Once agentic coding was a bust, I gave it a run as a daily driver for AI assistant. It performed fairly well but then began behaving strangely. It would lose context ", "context": "", "use_case": "", "switched_from": "", "story": "Gemini 3.1 Pro Preview"}, {"id": 754945681, "item_id": 47184105, "url": "https://news.ycombinator.com/item?id=47184105", "by": "idoxer", "date": "2026-02-27", "stance": "warn", "quote": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released", "excerpt": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released, now the Gemini 3 Flash is having peak times often. Also cause google to continuously reducing free tiers to stable their infrastructure And it's also seems to be worse from here, probably for all major providers", "context": "", "use_case": "", "switched_from": "", "story": "Gemini's 10 days of different outages and increasing high demand"}]}, {"label": "Tool calling", "summary": "It fails tool calls and gets stuck in loops in agentic coding tools.", "quotes": [{"id": 796666673, "item_id": 49791667, "url": "https://news.ycombinator.com/item?id=49791667", "by": "haellsigh", "date": "2026-09-21", "stance": "warn", "quote": "the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls", "excerpt": "I remember the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls. They really redeemed themselves with the 3.5 Flash series, it's great. I love how fast it is. Now if only they could offer a subscription service for it...", "context": "", "use_case": "text", "switched_from": "", "story": "Grok 4.7"}, {"id": 737029073, "item_id": 46064317, "url": "https://news.ycombinator.com/item?id=46064317", "by": "terrywang", "date": "2025-11-26", "stance": "warn", "quote": "relatively weak tool calling capability of the Gemini 2.5 Pro model", "excerpt": "Gemini CLI at this stage isn't good at complex coding tasks (vs. Claude Code, Codex, Cursor CLI, Qoder CLI, etc.). Mostly because of the simple ReAct loop, compounded by relatively weak tool calling capability of the Gemini 2.5 Pro model. > I haven't tried complex coding tasks using Gemini 3.0 Pro Preview yet. I reckon it won't be materially different. Gemini CLI is open source and being actively developed, which is ", "context": "", "use_case": "coding", "switched_from": "", "story": "Gemini CLI tips and tricks for agentic coding"}, {"id": 744602704, "item_id": 46537669, "url": "https://news.ycombinator.com/item?id=46537669", "by": "theturtletalks", "date": "2026-01-07", "stance": "warn", "quote": "Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem", "excerpt": "All of this seems like manufactured hype for Gemini. I use GPT-5.2, Opus 4.5, and Gemini 3 flash and pro with Droid CLI and Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem, and still fails to call tools consistently (sometimes the whole thread is corrupted and you can\u2019t rewind and use another model). Terminal Bench supports my findings, GPT-5.2 and", "context": "", "use_case": "agentic coding", "switched_from": "", "story": "Google Got Its Groove Back and Edged Ahead of OpenAI"}]}, {"label": "Cost and quotas", "summary": "It burns reasoning tokens and uses up usage limits quickly.", "quotes": [{"id": 753437728, "item_id": 47089858, "url": "https://news.ycombinator.com/item?id=47089858", "by": "startages", "date": "2026-02-20", "stance": "warn", "quote": "Gemini spends significantly more on reasoning tokens to produce lower quality answers", "excerpt": "This is misleading. I'm running a live experiment here: https://project80.divcrafts.com/ There are 4 models, all receiving the exact same prompts a few times a day, required to respond with a specific action. In the first experiment I used gemini-3-pro-preview, it spent ~$18 on the same task where Opus 4.5 spent ~$4, GPT-5.1 spent ~$4.50, and Grok spent ~$7. Pro was burning through money so fast I switched to gemini-", "context": "runs a live multi-model experiment", "use_case": "agent task experiment", "switched_from": "", "story": "Gemini 3.1 Pro Preview"}, {"id": 763837186, "item_id": 47739824, "url": "https://news.ycombinator.com/item?id=47739824", "by": "rzkyif", "date": "2026-04-12", "stance": "neutral", "quote": "I can burn through the Antigravity weekly limit in 1-2 hours if I force it to use Gemini 3.1 Pro.", "excerpt": "My personal experience is way different: I struggle to burn through more than 50% of the 5 hour limit For context, with Google AI Pro, I can burn through the Antigravity weekly limit in 1-2 hours if I force it to use Gemini 3.1 Pro. Meanwhile Gemini 3 Flash is basically unlimited but frequently produces buggy code or fail to implement things how I personally would (felt like it doesn't \"think\" like a software dev) I ", "context": "", "use_case": "coding", "switched_from": "", "story": "Pro Max 5x Quota Exhausted in 1.5 Hours Despite Moderate Usage"}, {"id": 761373057, "item_id": 47585816, "url": "https://news.ycombinator.com/item?id=47585816", "by": "fxtentacle", "date": "2026-03-31", "stance": "warn", "quote": "But it's not just Anthropic. I had the same issue with Gemini 3.1 Pro.", "excerpt": "Use AI! It'll make you a 10x engineer! (cost-wise) /s I've recently had the displeasure of Opus 4.6 hallucinating an API. It would have been great if that API had existed, but it did not. Still, it then looped until I manually terminated it while trying to make tests pass. In my case, I used up about $12 of usage in 30 minutes. My guess would be mostly through the (pretty verbose) thinking tokens. But it's not just A", "context": "", "use_case": "coding", "switched_from": "", "story": "Claude Code bug can silently 10-20x API costs"}]}, {"label": "Weaker than Opus", "summary": "Some find it less capable than Claude Opus for general use.", "quotes": [{"id": 804684144, "item_id": 50292759, "url": "https://news.ycombinator.com/item?id=50292759", "by": "oblio", "date": "2026-10-29", "stance": "warn", "quote": "Gemini is just... dumber overall.", "excerpt": "> Gemini is also IMO excellent for everything that is not coding. It's really rare that I use Anthropic models (which I pay a subscription for) for anything else than coding. I use Gemini Pro for every day stuff and Opus is much better than it. A lot more features and everything works better, especially in the websites. Gemini is just... dumber overall. I had a project/gem (now replaced by a skill) and sometimes I'd ", "context": "", "use_case": "everyday tasks", "switched_from": "", "story": "Google Search is now less than half of Alphabet\u2019s revenue"}]}, {"label": "Image editing", "summary": "One user found it did a poor job editing uploaded photos compared with Flash.", "quotes": [{"id": 735963073, "item_id": 45997692, "url": "https://news.ycombinator.com/item?id=45997692", "by": "Nemi", "date": "2025-11-20", "stance": "warn", "quote": "2.5 flash works really well but 2.5 pro or 3.0 pro do a very poor job", "excerpt": "I feel like I am going crazy or missed something simple but when I use the Gemini app and I ask it to edit a photo that I upload, 2.5 flash works really well but 2.5 pro or 3.0 pro do a very poor job. I uploaded an image of me and asked it to make me bald and flash did a great job of just changing me in the photo but 3.0 pro took me out of the photo completely and just created a headshot of a bald man that only sort ", "context": "", "use_case": "editing uploaded photos", "switched_from": "", "story": "Nano Banana Pro"}, {"id": 735963074, "item_id": 45997692, "url": "https://news.ycombinator.com/item?id=45997692", "by": "Nemi", "date": "2025-11-20", "stance": "warn", "quote": "2.5 flash works really well but 2.5 pro or 3.0 pro do a very poor job", "excerpt": "I feel like I am going crazy or missed something simple but when I use the Gemini app and I ask it to edit a photo that I upload, 2.5 flash works really well but 2.5 pro or 3.0 pro do a very poor job. I uploaded an image of me and asked it to make me bald and flash did a great job of just changing me in the photo but 3.0 pro took me out of the photo completely and just created a headshot of a bald man that only sort ", "context": "", "use_case": "editing uploaded photos", "switched_from": "", "story": "Nano Banana Pro"}]}, {"label": "Refusals", "summary": "It refused to run a security vulnerability-hunting task.", "quotes": [{"id": 784869635, "item_id": 49054352, "url": "https://news.ycombinator.com/item?id=49054352", "by": "alexsmirnov", "date": "2026-07-25", "stance": "warn", "quote": "gpt-5.6, gemini pro, and opus all rejected to follow.", "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"}]}], "who": [{"by": "toshvelaga", "context": "built a healthcare LLM benchmark", "url": "https://news.ycombinator.com/item?id=49631768", "date": "2026-09-09"}, {"by": "ryptophan", "context": "Tech Lead at Georgian", "url": "https://news.ycombinator.com/item?id=49550977", "date": "2026-09-03"}, {"by": "glimshe", "context": "Pro subscriber without a corporation behind them", "url": "https://news.ycombinator.com/item?id=48993529", "date": "2026-07-21"}, {"by": "radio879", "context": "works with LLMs daily for 8-15 hours", "url": "https://news.ycombinator.com/item?id=47318434", "date": "2026-03-09"}, {"by": "adrihd", "context": "has ADHD", "url": "https://news.ycombinator.com/item?id=46985313", "date": "2026-02-11"}, {"by": "erichocean", "context": "has AI write ~100% of code, Clojure/Java", "url": "https://news.ycombinator.com/item?id=46719202", "date": "2026-01-22"}, {"by": "hardmaru", "context": "Sakana AI team", "url": "https://news.ycombinator.com/item?id=46500090", "date": "2026-01-05"}], "tags": ["for everyday questions", "for code refactoring", "for long-context work", "for planning", "for UI polish"], "use_cases": ["coding", "ai coding", "editing uploaded photos", "everyday tasks", "chat and writing", "troubleshooting logic and finding bugs"], "compared": [{"slug": "gemini-flash", "name": "Gemini Flash", "co": 23, "switched_from": 2, "switched_to": 2}, {"slug": "claude-opus", "name": "Claude Opus", "co": 16, "switched_from": 0, "switched_to": 1}, {"slug": "gpt-5", "name": "GPT-5", "co": 15, "switched_from": 1, "switched_to": 0}, {"slug": "codex", "name": "Codex", "co": 11, "switched_from": 0, "switched_to": 1}, {"slug": "claude-code", "name": "Claude Code", "co": 11, "switched_from": 0, "switched_to": 0}, {"slug": "antigravity", "name": "Antigravity", "co": 7, "switched_from": 0, "switched_to": 0}, {"slug": "claude-sonnet", "name": "Claude Sonnet", "co": 6, "switched_from": 0, "switched_to": 0}, {"slug": "glm", "name": "GLM", "co": 6, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47074735", "title": "Gemini 3.1 Pro Preview", "mentions": 4}, {"url": "https://news.ycombinator.com/item?id=49913571", "title": "Gemini 4 Argon", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49788838", "title": "Grok 4.7", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=49289112", "title": "Gemini 3.7 Flash", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=46626639", "title": "Ask HN: What is best way to provide continuous context to models?", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=46528389", "title": "Google Got Its Groove Back and Edged Ahead of OpenAI", "mentions": 2}], "sentiment_by_month": {"2025-11": {"neutral": 1, "compare": 2, "recommend": 3, "warn": 2}, "2025-12": {"neutral": 3, "compare": 1, "recommend": 4}, "2026-01": {"recommend": 5, "warn": 1, "neutral": 3}, "2026-02": {"neutral": 2, "recommend": 3, "warn": 3}, "2026-03": {"recommend": 2, "neutral": 2, "warn": 1}, "2026-04": {"neutral": 4, "recommend": 1}, "2026-05": {"recommend": 2}, "2026-06": {"neutral": 3, "recommend": 2}, "2026-07": {"recommend": 1, "neutral": 3, "warn": 1}, "2026-08": {"neutral": 3}, "2026-09": {"neutral": 3, "recommend": 5, "warn": 1}, "2026-10": {"recommend": 2, "warn": 1}}, "quotes": [{"id": 804684144, "item_id": 50292759, "url": "https://news.ycombinator.com/item?id=50292759", "by": "oblio", "date": "2026-10-29", "stance": "warn", "quote": "Gemini is just... dumber overall.", "excerpt": "> Gemini is also IMO excellent for everything that is not coding. It's really rare that I use Anthropic models (which I pay a subscription for) for anything else than coding. I use Gemini Pro for every day stuff and Opus is much better than it. A lot more features and everything works better, especially in the websites. Gemini is just... dumber overall. I had a project/gem (now replaced by a skill) and sometimes I'd ", "context": "", "use_case": "everyday tasks", "switched_from": "", "story": "Google Search is now less than half of Alphabet\u2019s revenue"}, {"id": 804462736, "item_id": 50278921, "url": "https://news.ycombinator.com/item?id=50278921", "by": "5555watch", "date": "2026-10-28", "stance": "recommend", "quote": "I loved Gemini 3.1 Pro", "excerpt": "I loved Gemini 3.1 Pro and it was very sad when they nerfed it at around May this year. I think it did pretty good as a chat, and quite decent on Antigravity (tried very little). Been sitting on 5.6 / Astra and Codex since. Occasionally do bring up the 3.1 Pro through the API, because I love the writing style, but keep noticing that I'm much more productive with the frontier, despite the love to Gemini. Tried 4 Argon", "context": "", "use_case": "chat and writing", "switched_from": "", "story": "Gemini 4 Carbon"}, {"id": 798662544, "item_id": 49916409, "url": "https://news.ycombinator.com/item?id=49916409", "by": "landdate", "date": "2026-10-01", "stance": "recommend", "quote": "Its easily the best model I have tried. And yes I am talking about 3.1 pro.", "excerpt": "I don't like AI, but it's very enjoyable for me to see my predictions on the success of gemini come to fruition. I only use gemini, and while I don't use it for actually writing up code, I use it to help me troibleshoot my logic and help find bugs. Its easily the best model I have tried. And yes I am talking about 3.1 pro. Also I have found gemini the only model to be the least likely to douse me in flattery, and wil", "context": "", "use_case": "troubleshooting logic and finding bugs", "switched_from": "", "story": "Gemini 4 Argon"}, {"id": 798636401, "item_id": 49914775, "url": "https://news.ycombinator.com/item?id=49914775", "by": "p_l", "date": "2026-09-30", "stance": "neutral", "quote": "mine has 3.1 Pro, 3.5 Flash-Lite, and 3.8 Flash (with optional extra effort)", "excerpt": "Not sure what's the problem with phone app rollout, mine has 3.1 Pro, 3.5 Flash-Lite, and 3.8 Flash (with optional extra effort), and pretty much it switched to latest Flash series as main option since 3.5 just after release", "context": "", "use_case": "phone app model", "switched_from": "", "story": "Gemini 4 Argon"}, {"id": 797394784, "item_id": 49837174, "url": "https://news.ycombinator.com/item?id=49837174", "by": "hypfer", "date": "2026-09-24", "stance": "recommend", "quote": "Thank you, Gemini. Much appreciated.", "excerpt": "If you allow me to quote Gemini 3.1 pro: ___ If we are talking about the standard 3.5-inch High Density (HD) floppy disk, which is the universal cultural icon for \"floppy disk\" (and the save icon in your software), the capacity is exactly: 1,474,560 bytes This is equal to 1,440 KB (using 1024-byte kilobytes). It is marketed and universally referred to as 1.44 MB. Now, let's run the exact math on DHH's keynote claim: ", "context": "", "use_case": "answering a factual question", "switched_from": "", "story": "Rails World 2026 Opening Keynote [video]"}, {"id": 796666673, "item_id": 49791667, "url": "https://news.ycombinator.com/item?id=49791667", "by": "haellsigh", "date": "2026-09-21", "stance": "warn", "quote": "the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls", "excerpt": "I remember the Gemini 2.5 Pro series was the best at text, but failed half of it's tool calls. They really redeemed themselves with the 3.5 Flash series, it's great. I love how fast it is. Now if only they could offer a subscription service for it...", "context": "", "use_case": "text", "switched_from": "", "story": "Grok 4.7"}, {"id": 796659840, "item_id": 49791240, "url": "https://news.ycombinator.com/item?id=49791240", "by": "moojacob", "date": "2026-09-21", "stance": "recommend", "quote": "I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff.", "excerpt": "I'll have to try Gemini Flash for coding. The reason I haven't I used Gemini for coding is last time I tried it couldn't call tools very well. I am a huge fan of Gemini Pro for chat... gemini somehow just knows the most obscure stuff. I'll double check something Gemini said and find the source is deep inside a hard to access scientific paper. Google just has the best index of the internet.", "context": "", "use_case": "chat and research", "switched_from": "", "story": "Grok 4.7"}, {"id": 794825492, "item_id": 49676593, "url": "https://news.ycombinator.com/item?id=49676593", "by": "karmakaze", "date": "2026-09-12", "stance": "neutral", "quote": "sometimes a GPT or Gemini 3.1 Pro", "excerpt": "Personally I'm using Qwen3.8-27B (MXFP4 quant W4A8) locally hosted on a pair of AMD GPUs (with DeepSeek Harness). It starts at 250 tokens/sec down to 120 past 128k context. At work mostly Opus 4.8 (sometimes a GPT or Gemini 3.1 Pro). I find Opus 5 chatty/slower and Fable can venture into over-engineering itself into unnecessary complications.", "context": "", "use_case": "work", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 794703424, "item_id": 49668964, "url": "https://news.ycombinator.com/item?id=49668964", "by": "grahamnorton39", "date": "2026-09-12", "stance": "recommend", "quote": "2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text", "excerpt": "Might be missing something \u2014- are there any issues with Gemini 3.1 Pro that aren\u2019t there in 2.5? I agree, though. 2.5 Pro is a great model. Very competent, knows a lot, and can process tons of text (and videos, and images, and audio too iirc?). Basically unlimited access to it too via AI Studio. I used it for processing and transforming bucketloads of data, ingesting masses of transcripts and converting them to flash", "context": "", "use_case": "processing transcripts into flashcards", "switched_from": "", "story": "Ask HN: Did Google kill its enterprise workhorse model?"}, {"id": 794108289, "item_id": 49631768, "url": "https://news.ycombinator.com/item?id=49631768", "by": "toshvelaga", "date": "2026-09-09", "stance": "recommend", "quote": "Gemini 3.8 flash and gemini 3.1 pro seem to lead across the other LLMs.", "excerpt": "There are no sites like artificialanalysis.ai or arena.ai for healthcare benchmarks so I decided to build it. I ran the leading models from openai, anthropic, and google on MedXpertQA: https://github.com/TsinghuaC3I/MedXpertQA#leaderboard The results show scores across organ systems and across diagnosis, treatment, and basic science. Gemini 3.8 flash and gemini 3.1 pro seem to lead across the other LLMs. The last tim", "context": "built a healthcare LLM benchmark", "use_case": "medical question benchmark", "switched_from": "", "story": "Benchmarked how the latest LLMs perform on expert level medical questions"}, {"id": 793939905, "item_id": 49621244, "url": "https://news.ycombinator.com/item?id=49621244", "by": "Rodmine", "date": "2026-09-09", "stance": "recommend", "quote": "I switch back to 3.1 pro and it works fine.", "excerpt": "Google definitely does not have good models and if it does, Gemini Flash series models are definitely not them. I use the Google Gemini app daily (for searches, research etc.), everyday, for some query, gemini flash 3.8 hallucinates, does not understand context etc. in quite short conversations. I switch back to 3.1 pro and it works fine. Which brings me to why do they get good benchmark scores? Have never really und", "context": "", "use_case": "searches and research", "switched_from": "Gemini Flash 3.8", "story": "Muse \u2013 Meta\u2019s personal AI agent"}, {"id": 792815632, "item_id": 49550977, "url": "https://news.ycombinator.com/item?id=49550977", "by": "ryptophan", "date": "2026-09-03", "stance": "neutral", "quote": "gemini-2.5-pro, claude-opus-4.8, gpt-5.1, llama-3.3-70b", "excerpt": "Author here. I'm a Tech Lead at Georgian (a growth-stage VC). Short version: I used a few retrieval benchmark datasets, and experimented with a 21 embedding model leaderboard. I held the benchmark corpus and relevance labels fixed, and transformed the queries: paraphrase, terse, verbose. I kept the \u201cinformation need\u201d the same under each query transformation, but the leaderboard order changed more than I expected. I r", "context": "Tech Lead at Georgian", "use_case": "rewriting benchmark queries", "switched_from": "", "story": "Your Queries Can Reorder the Retrieval Leaderboard"}, {"id": 788655041, "item_id": 49290940, "url": "https://news.ycombinator.com/item?id=49290940", "by": "dismalaf", "date": "2026-08-13", "stance": "neutral", "quote": "I'll escalate to Gemini Pro with extended thinking", "excerpt": "Yup. I use it for a ton of mundane queries (stuff that I might have used Google search for in the past) and it's great. Nice and fast and correct more often than not, especially if you prompt it in a way that it invokes Google search (but filters out ads and SEO slop). It's even alright at programming tasks but if it stumbles then I'll escalate to Gemini Pro with extended thinking.", "context": "", "use_case": "harder programming tasks", "switched_from": "", "story": "Gemini 3.7 Flash"}, {"id": 788638721, "item_id": 49289920, "url": "https://news.ycombinator.com/item?id=49289920", "by": "eckr", "date": "2026-08-13", "stance": "neutral", "quote": "whenever I use Gemini I've just been using 3.1 Pro", "excerpt": "Maybe this is just my experience, but have people had trouble with 3.6 Flash just... getting things it has seen in its context correct? I don't know if it's been insanely benchmaxxed or what, but it'll pull information from websites and immediately get it wrong the token after. Or for example (this is something that happened like yesterday) I asked it to compare the uses of A and B in a language I was learning, and t", "context": "", "use_case": "", "switched_from": "Gemini 3.6 Flash", "story": "Gemini 3.7 Flash"}, {"id": 787697424, "item_id": 49231089, "url": "https://news.ycombinator.com/item?id=49231089", "by": "gste", "date": "2026-08-09", "stance": "neutral", "quote": "I used the Pro subscription for Gemini 3 Pro", "excerpt": "I used the Pro subscription for Gemini 3 Pro, which I think is running Nano Banana Pro for image generation, but it doesn't tell you which model in the chatbot UI.", "context": "", "use_case": "image generation", "switched_from": "", "story": "Gemini 3 Pro can generate normal maps"}, {"id": 784869635, "item_id": 49054352, "url": "https://news.ycombinator.com/item?id=49054352", "by": "alexsmirnov", "date": "2026-07-25", "stance": "warn", "quote": "gpt-5.6, gemini pro, and opus all rejected to follow.", "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": 783896464, "item_id": 48993529, "url": "https://news.ycombinator.com/item?id=48993529", "by": "glimshe", "date": "2026-07-21", "stance": "recommend", "quote": "I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review", "excerpt": "I respectfully disagree, at least for most tasks. 3.1 Pro: while it's coding performance is mediocre, a lot of coding work requires minimum actual thinking. I use it often for light refactoring, boilerplate generation, testcase skeleton generation, code review, language questions (\"is there a better way to write this code block?\"). I don't have a corporation behind me so costs matter. Considering that I'm a Pro subsc", "context": "Pro subscriber without a corporation behind them", "use_case": "light refactoring and boilerplate", "switched_from": "", "story": "Gemini 3.6 Flash"}, {"id": 783707040, "item_id": 48981690, "url": "https://news.ycombinator.com/item?id=48981690", "by": "ItsBob", "date": "2026-07-20", "stance": "neutral", "quote": "I use Gemini Pro (got it with my 5TB of Google storage)", "excerpt": "It's not losing yet but I think it will. I use Gemini Pro (got it with my 5TB of Google storage) and for a while it seemed if Google had pulled the rug as I was running out of quota after only a few hours. That seems to have been dialled back a bit lately... I also use Chatbot with Deepseek V4 Pro and GLM 5.2. However, GLM 5.2 seems to eat tokens like crazy as the context increases. Anyway, there isn't a meaningful e", "context": "", "use_case": "general AI assistant", "switched_from": "", "story": "American AI is locked down and proprietary. It's losing"}, {"id": 783628816, "item_id": 48976801, "url": "https://news.ycombinator.com/item?id=48976801", "by": "glimshe", "date": "2026-07-20", "stance": "neutral", "quote": "I told Gemini Pro I woke up after having dreamt that polynomial", "excerpt": "I told Gemini Pro I woke up after having dreamt that polynomial and asked if it was related to the Jacobian conjecture. It spent some time thinking and referenced this tweet announcement, saying: \"If you truly dreamt about that specific polynomial, you might be mathematically clairvoyant.\" In the rest of the answer, it maintained a cautious skepticism about my claim, saying: \"Here is exactly why the math world is cur", "context": "", "use_case": "discussing math conjecture", "switched_from": "", "story": "Claude Fable produced a counterexample to the Jacobian Conjecture"}, {"id": 782553153, "item_id": 48909572, "url": "https://news.ycombinator.com/item?id=48909572", "by": "jchw", "date": "2026-07-14", "stance": "neutral", "quote": "in a conversation with Gemini 3 Pro", "excerpt": "I think it might be even worse. LLMs seem to get tragically stuck on certain patterns. Maybe it's partly because a pile of weights essentially always starts from scratch in the same condition, but even within a single conversation, it will literally just latch onto words and repeat them incessantly, to the point where it becomes annoying. So for example, current Claude models love \"honest\". They are always producing ", "context": "", "use_case": "", "switched_from": "", "story": "How to stop Claude from saying load-bearing"}, {"id": 780134930, "item_id": 48758433, "url": "https://news.ycombinator.com/item?id=48758433", "by": "muzani", "date": "2026-07-02", "stance": "recommend", "quote": "Gemini 3 Pro is actually at a mid designer level.", "excerpt": "In the last 12 months? Antigravity started the workflow where you give a list of things to do and it goes off and does those things without supervision, including drafting and testing edge cases. It can even spin up images. Fable is the latest form of this workflow. Gemini 3 Pro is actually at a mid designer level. None of the others are even at a junior human design level. Sonnet 4.5 was, for a brief moment, creativ", "context": "", "use_case": "design", "switched_from": "", "story": "Ask HN: Does anybody still FEEL improvements between latest LLMs for coding?"}, {"id": 778370786, "item_id": 48648174, "url": "https://news.ycombinator.com/item?id=48648174", "by": "cromka", "date": "2026-06-23", "stance": "neutral", "quote": "now also have access Gemini Pro and Claude Pro", "excerpt": "I use Go plan precisely with Opencode IDE (and also Jetbrains IDE suite), but now also have access Gemini Pro and Claude Pro. And wonder which tooling to invest my time into, especially that MCP servers also potentially come into play here, and I want at least some models/tools to handle private tasks, like handling my increasingly-complex Home Assistant setup. And I also want to start using models according to needs", "context": "", "use_case": "AI models", "switched_from": "", "story": "Elevated error rate across multiple models"}, {"id": 777648163, "item_id": 48603010, "url": "https://news.ycombinator.com/item?id=48603010", "by": "joe_mamba", "date": "2026-06-19", "stance": "recommend", "quote": "That's what the 3.1 Pro is better for.", "excerpt": ">from the looks of it, 3.5 Flash is still better than most models I guess it depends on what you're using it for. I think it's absolute garbage for coding and reasoning. I use it almost daily as an alternative to google search, but for questions related to coding, solving arch linux and wine lutris issues, helping me with MXLinux issues, and wifi issues on an old rooted huawei tablet running LineageOS, it was consist", "context": "", "use_case": "complex issues", "switched_from": "", "story": "John Jumper leaves Google to join Anthropic"}, {"id": 776733780, "item_id": 48545861, "url": "https://news.ycombinator.com/item?id=48545861", "by": "mark_l_watson", "date": "2026-06-15", "stance": "neutral", "quote": "I use Opus + Gemini Pro cumulatively for 3 or 4 hours a week.", "excerpt": "I would like to say I run 100% local, but I use Opus + Gemini Pro cumulatively for 3 or 4 hours a week. I also like to use DeepSeek v4 flash with OpenCode for small quick tasks. I did just publish a free to read online book \"The Rise of Local Coding Agents\" [1] where I document my setup that I enjoy using. I use little-coder (built on pi) and have good results for small Python and TypeScript applications. I struggle ", "context": "", "use_case": "coding", "switched_from": "", "story": "Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?"}, {"id": 775256032, "item_id": 48453502, "url": "https://news.ycombinator.com/item?id=48453502", "by": "novafunc", "date": "2026-06-08", "stance": "recommend", "quote": "I used Gemini Pro to convert health data from a .csv and import it into Apple Health. Worked first try.", "excerpt": "I used Gemini Pro to convert health data from a .csv and import it into Apple Health. Worked first try. I just hope the AI Apple use is smart enough for the task. And that giving the AI information it needs is easy (such as a snippet of the health data so it knows what it's working with).", "context": "", "use_case": "converting health data to Apple Health", "switched_from": "", "story": "Apple WWDC 2026 Livestream"}, {"id": 774511889, "item_id": 48406993, "url": "https://news.ycombinator.com/item?id=48406993", "by": "brianjmingus", "date": "2026-06-04", "stance": "neutral", "quote": "I used Antigravity CLI with gemini-3.1-pro-preview.", "excerpt": "It's roughly the state of the art. I developed it over the last day. I used Antigravity CLI with gemini-3.1-pro-preview.", "context": "", "use_case": "developing the project", "switched_from": "", "story": "Do transformers need three projections? Systematic study of QKV variants"}, {"id": 769812528, "item_id": 48113283, "url": "https://news.ycombinator.com/item?id=48113283", "by": "xiphias2", "date": "2026-05-12", "stance": "recommend", "quote": "I love using the Gemini pro model for quick questions", "excerpt": "Google needs to beat OpenAI and Antropic in coding models because that's where the big money is going. I love using the Gemini pro model for quick questions, but that's not where I'm spending the real money. They have so many great software engineers but unable to use them to speed up coding AI research. Hopefully with Sergey's focus it will get better. This cursor thing is just another experiment nobody cares about.", "context": "", "use_case": "quick questions", "switched_from": "", "story": "AI Pointer from DeepMind"}, {"id": 769435265, "item_id": 48089704, "url": "https://news.ycombinator.com/item?id=48089704", "by": "simonpure", "date": "2026-05-10", "stance": "recommend", "quote": "I've noticed a big jump when they switched to Gemini 3.1 Pro and it really became useful.", "excerpt": "Thanks! I've noticed a big jump when they switched to Gemini 3.1 Pro and it really became useful. I like that I can use it from my phone too. It took a bit of trial and error but I came up with a good ralph loop between GitHub Actions and Google Jules using the Jules API. So basically I have Jules extend its TODO.md with the next set of tasks and open a PR then run a GitHub Action with a few checks, auto-merge, and t", "context": "", "use_case": "powering Jules", "switched_from": "", "story": "Ask HN: What Are You Working On? (May 2026)"}, {"id": 766737347, "item_id": 47921084, "url": "https://news.ycombinator.com/item?id=47921084", "by": "mark_l_watson", "date": "2026-04-27", "stance": "neutral", "quote": "fun using strong AIs like Claude and Gemini Pro in AntiGravity", "excerpt": "Indirectly I use Mistral daily: I like Proton\u2019s Lumo private chat and that runs on Mistral technology. Something like Lumo is very much good enough to replace search and general information browsing and for me is very practical. What is not so practical is my paying for Gemini Ultra, which has some practicality but is something I pay for because it is fun using strong AIs like Claude and Gemini Pro in AntiGravity. It", "context": "", "use_case": "", "switched_from": "", "story": "France's Mistral Built a $14B AI Empire by Not Being American"}, {"id": 765528194, "item_id": 47845512, "url": "https://news.ycombinator.com/item?id=47845512", "by": "0xbadcafebee", "date": "2026-04-21", "stance": "neutral", "quote": "I use GPT-5.4, GPT-5.3-Codex, Gemini-3-Pro, Opus, Sonnet, at work every week.", "excerpt": "A model cannot read your mind. It can guess, and those guesses are more likely to be wrong if you don't give it the right input, and model performance gets worse if not steered/curated properly. The output depends on the input. https://medium.com/@adambaitch/the-model-vs-the-harness-which-actually-matters-more-59dd3116bb31 | https://aakashgupta.medium.com/2025-was-agents-2026-is-agent-harnesses-heres-why-that-changes", "context": "", "use_case": "work coding", "switched_from": "", "story": "Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving"}, {"id": 763837186, "item_id": 47739824, "url": "https://news.ycombinator.com/item?id=47739824", "by": "rzkyif", "date": "2026-04-12", "stance": "neutral", "quote": "I can burn through the Antigravity weekly limit in 1-2 hours if I force it to use Gemini 3.1 Pro.", "excerpt": "My personal experience is way different: I struggle to burn through more than 50% of the 5 hour limit For context, with Google AI Pro, I can burn through the Antigravity weekly limit in 1-2 hours if I force it to use Gemini 3.1 Pro. Meanwhile Gemini 3 Flash is basically unlimited but frequently produces buggy code or fail to implement things how I personally would (felt like it doesn't \"think\" like a software dev) I ", "context": "", "use_case": "coding", "switched_from": "", "story": "Pro Max 5x Quota Exhausted in 1.5 Hours Despite Moderate Usage"}, {"id": 763364706, "item_id": 47710294, "url": "https://news.ycombinator.com/item?id=47710294", "by": "dokdev", "date": "2026-04-09", "stance": "recommend", "quote": "If I need some UI polish, I switch to gemini 3.1 pro.", "excerpt": "Cursor $20 subscription is also very good. auto is usually decent. I use composer 2 for heavy tasks and it does good job. If I need some UI polish, I switch to gemini 3.1 pro. The trick is to stay at auto/composer 2. But if you try to use other premium models too much, then you\u2019ll hit the limit as well.", "context": "", "use_case": "UI polish", "switched_from": "", "story": "Claude Code limits are starting to feel like a psychological trick"}, {"id": 762956451, "item_id": 47684778, "url": "https://news.ycombinator.com/item?id=47684778", "by": "SilverSlash", "date": "2026-04-07", "stance": "neutral", "quote": "Gemini 3.1 Pro on High have been able to solve a very straightforward and basic UI bug", "excerpt": "Someone needs to make a compilation of all these classic OpenAI moments. Including hits like GPT-2 too dangerous, the 64x64 image model DALL-E too scary, \"push the veil of ignorance back\", AGI achieved internally, Q*/strawberry is able to solve math and is making OpenAI researchers panic, etc. etc. I use Codex btw, and I really love it. But some of these companies have been so overhyping the capabilities of these mod", "context": "", "use_case": "debugging a UI bug", "switched_from": "", "story": "OpenAI says its new model GPT-2 is too dangerous to release (2019)"}, {"id": 761373057, "item_id": 47585816, "url": "https://news.ycombinator.com/item?id=47585816", "by": "fxtentacle", "date": "2026-03-31", "stance": "warn", "quote": "But it's not just Anthropic. I had the same issue with Gemini 3.1 Pro.", "excerpt": "Use AI! It'll make you a 10x engineer! (cost-wise) /s I've recently had the displeasure of Opus 4.6 hallucinating an API. It would have been great if that API had existed, but it did not. Still, it then looped until I manually terminated it while trying to make tests pass. In my case, I used up about $12 of usage in 30 minutes. My guess would be mostly through the (pretty verbose) thinking tokens. But it's not just A", "context": "", "use_case": "coding", "switched_from": "", "story": "Claude Code bug can silently 10-20x API costs"}, {"id": 760202211, "item_id": 47512638, "url": "https://news.ycombinator.com/item?id=47512638", "by": "Patt_", "date": "2026-03-24", "stance": "neutral", "quote": "They have free allowances of Claude and Gemini 3.1 Pro there", "excerpt": "It's all over Chinese Taobao. biggest catch would be that all of the sutff is in Chinese, they have an international version in english but if you run into issues and have to talk to the seller, 99% of them don't speak Chinese but with AI translation that won't be a problem. you can even get Claude Code shared plan there as well, what they do is to buy the MAX plan and then have a couple people sharing it, which on a", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Ask HN: $50 monthly budget, which coding models would you recommend now?"}, {"id": 758053761, "item_id": 47378360, "url": "https://news.ycombinator.com/item?id=47378360", "by": "jeremychone", "date": "2026-03-14", "stance": "neutral", "quote": "Gemini Pro and Flash have had 1M context for a long time", "excerpt": "Interesting, I\u2019ve never needed 1M, or even 250k+ context. I\u2019m usually under 100k per request. About 80% of my code is AI-generated, with a controlled workflow using dev-chat.md and spec.md. I use Flash for code maps and auto-context, and GPT-4.5 or Opus for coding, all via API with a custom tool. Gemini Pro and Flash have had 1M context for a long time, but even though I use Flash 3 a lot, and it\u2019s awesome, I\u2019ve neve", "context": "", "use_case": "", "switched_from": "", "story": "1M context is now generally available for Opus 4.6 and Sonnet 4.6"}, {"id": 757589697, "item_id": 47349356, "url": "https://news.ycombinator.com/item?id=47349356", "by": "stared", "date": "2026-03-12", "stance": "recommend", "quote": "My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up.", "excerpt": "I spend much more time refactoring that creating features (though, it is getting better with each model). My go-to approach is to use Claude Code Opus 4.6 for writing and Gemini 3.1 Pro for cleaning up. I feel that doing it just one-stage is rarely enough. A lot of prompts about finding the right level of abstraction, DRY, etc. An earlier example (Opus 4.5 + Gemini 3 Pro) is here: https://github.com/stared/sc2-balanc", "context": "", "use_case": "refactoring and cleanup", "switched_from": "", "story": "Many SWE-bench-Passing PRs would not be merged"}, {"id": 757094946, "item_id": 47318434, "url": "https://news.ycombinator.com/item?id=47318434", "by": "radio879", "date": "2026-03-09", "stance": "recommend", "quote": "use pro for planning, fixing, problem solving, all the difficult stuff", "excerpt": "With all of the talk about \"LLMs are non deterministic\" I am trying to build my own tool (at this exact moment, since I seem to have gotten really good at this) which does some of what these OpenClaw/Clawdbots do for people, but instead of having these LLMs/agents handle everything I am trying to do as much of it as pure deterministic code as possible. I test things once, like daily automations, and they will repeat ", "context": "works with LLMs daily for 8-15 hours", "use_case": "planning and problem solving", "switched_from": "", "story": "Ask HN: Read\u2011only LLM tool for email triage and knowledge extraction?"}, {"id": 754945681, "item_id": 47184105, "url": "https://news.ycombinator.com/item?id=47184105", "by": "idoxer", "date": "2026-02-27", "stance": "warn", "quote": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released", "excerpt": "Getting worse each day, it was mainly Gemini 3 Pro, switched to 3.1 when it released, now the Gemini 3 Flash is having peak times often. Also cause google to continuously reducing free tiers to stable their infrastructure And it's also seems to be worse from here, probably for all major providers", "context": "", "use_case": "", "switched_from": "", "story": "Gemini's 10 days of different outages and increasing high demand"}, {"id": 753716576, "item_id": 47107286, "url": "https://news.ycombinator.com/item?id=47107286", "by": "brandall10", "date": "2026-02-21", "stance": "recommend", "quote": "I have a planner skill I use w/ Gemini Pro to generate new plans.", "excerpt": "I go a bit further than this and have had incredible success with 3 doc types: - Specs: these are generally static, but updatable as the project evolves. And they're broken out to an index file that gives a project overview, a high-level arch file, and files for all the main modules. Roughly ~1k lines of spec for 10k lines of code - Plans: these are the output of a planning session with an LLM. They point to the asso", "context": "", "use_case": "generating plans", "switched_from": "", "story": "How I use Claude Code: Separation of planning and execution"}, {"id": 753690310, "item_id": 47105644, "url": "https://news.ycombinator.com/item?id=47105644", "by": "dudeinhawaii", "date": "2026-02-21", "stance": "warn", "quote": "Gemini 3 Pro Preview went from exceptional in the first month to mediocre", "excerpt": "After 2 days of giving it a go, I find that Gemini CLI is still considerably worse than both Codex and Claude Code. The model itself also has strange behaviors that seem like it gets randomly replaced with Gemini-3-Flash or something else. I'll explain. Once agentic coding was a bust, I gave it a run as a daily driver for AI assistant. It performed fairly well but then began behaving strangely. It would lose context ", "context": "", "use_case": "", "switched_from": "", "story": "Gemini 3.1 Pro Preview"}, {"id": 744602704, "item_id": 46537669, "url": "https://news.ycombinator.com/item?id=46537669", "by": "theturtletalks", "date": "2026-01-07", "stance": "warn", "quote": "Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem", "excerpt": "All of this seems like manufactured hype for Gemini. I use GPT-5.2, Opus 4.5, and Gemini 3 flash and pro with Droid CLI and Gemini is consistently the worst. It gets stuck in loops, wants to wipe projects when it can\u2019t figure out the problem, and still fails to call tools consistently (sometimes the whole thread is corrupted and you can\u2019t rewind and use another model). Terminal Bench supports my findings, GPT-5.2 and", "context": "", "use_case": "agentic coding", "switched_from": "", "story": "Google Got Its Groove Back and Edged Ahead of OpenAI"}, {"id": 737029073, "item_id": 46064317, "url": "https://news.ycombinator.com/item?id=46064317", "by": "terrywang", "date": "2025-11-26", "stance": "warn", "quote": "relatively weak tool calling capability of the Gemini 2.5 Pro model", "excerpt": "Gemini CLI at this stage isn't good at complex coding tasks (vs. Claude Code, Codex, Cursor CLI, Qoder CLI, etc.). Mostly because of the simple ReAct loop, compounded by relatively weak tool calling capability of the Gemini 2.5 Pro model. > I haven't tried complex coding tasks using Gemini 3.0 Pro Preview yet. I reckon it won't be materially different. Gemini CLI is open source and being actively developed, which is ", "context": "", "use_case": "coding", "switched_from": "", "story": "Gemini CLI tips and tricks for agentic coding"}]}