{"slug": "gpt-5-thinking", "name": "GPT-5 Thinking", "category": "ai", "category_name": "AI models and assistants", "description": "", "recommenders": 6, "commenters": 14, "warners": 2, "score": 4.806, "rising": -1.96, "first": "2025-11-03", "latest": "2026-04-09", "why": "I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly.", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 7, "neutral": 6, "compare": 0, "warn": 2}, "spark": [2, 1, 2, 0, 1, 1, 0, 0, 0, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly.", "kind": "", "icon": false, "domain": "chatgpt.com", "github_repo": null, "official_url": "", "pricing": "", "aliases": ["GPT-5.4 Thinking", "GPT-5.1 Thinking", "GPT-5.2 Thinking", "GPT-5 Thinking", "GPT-5.2-Thinking"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 763108432, "item_id": 47694277, "url": "https://news.ycombinator.com/item?id=47694277", "by": "simianwords", "date": "2026-04-08", "stance": "recommend", "quote": "It doesn't bullshit on the GPT-5.4 thinking version.", "excerpt": "I wanted this challenge with the thinking version (I apologised for it and edited the earlier version). It doesn't bullshit on the GPT-5.4 thinking version. Here is the result with thinking https://chatgpt.com/share/69d69dd6-fb50-838d-863c-4e1eda5d088b I suggest you try it yourself to be convinced. Try it in incognito mode if you wish. Or not.", "context": "", "use_case": "", "switched_from": "", "story": "The Future of Everything Is Lies, I Guess"}, {"id": 756362882, "item_id": 47272680, "url": "https://news.ycombinator.com/item?id=47272680", "by": "andy12_", "date": "2026-03-06", "stance": "recommend", "quote": "It concisely replies \"GPT-5.4 Thinking\".", "excerpt": "It's not a rumor; you can just test it. Ask the router \"What model are you\". It will yap on and on about being a GPT-5.3 model (Non-thinking models of OpenAI are insufferable yappers that don't know when to shut up). Ask it now \"What model are you. Think carefully\". It concisely replies \"GPT-5.4 Thinking\". https://openai.com/index/introducing-gpt-5/ > GPT\u20115 is a unified system with a smart, efficient model that answe", "context": "", "use_case": "", "switched_from": "", "story": "GPT-5.4 Thinking System Card"}, {"id": 746493264, "item_id": 46655829, "url": "https://news.ycombinator.com/item?id=46655829", "by": "timpera", "date": "2026-01-16", "stance": "recommend", "quote": "I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly.", "excerpt": "People in the comments are all talking about a degradation in performance, but I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly. There's no way to know for sure if the model is experiencing issues, that's the problem with this non-deterministic tech.", "context": "", "use_case": "searches", "switched_from": "", "story": "GPT-5.2 does not follow instructions and ignores my prompts"}, {"id": 745630641, "item_id": 46601915, "url": "https://news.ycombinator.com/item?id=46601915", "by": "andy12_", "date": "2026-01-13", "stance": "recommend", "quote": "When I use GPT 5.2-thinking for my ML research job, I pretty much never encounter hallucinations.", "excerpt": "> I don't entirely understand your extreme optimism towards LLMs given this proclivity for hallucination Simply because I don't see hallucinations as a permanent problem. I see that models keep improving more and more in this regard, and I don't see why the hallucination rate can't be abirtrarily reduced with further improvements to the architecture. When I ask Claude about obscure topics, it correctly replies \"I don", "context": "ML researcher", "use_case": "ML research", "switched_from": "", "story": "TimeCapsuleLLM: LLM trained only on data from 1800-1875"}, {"id": 738480465, "item_id": 46155029, "url": "https://news.ycombinator.com/item?id=46155029", "by": "XenophileJKO", "date": "2025-12-04", "stance": "recommend", "quote": "I started with o3 and now I'm using 5.1 Thinking (set to max)", "excerpt": "So.. I have been using an LLM to make 30 day buy and hold portfolios. And the results are \"ok\". (Like 8% vs 6% for the S&P 500 over the last 90 days) What you ask the model to do is super important. Just like writing or coding.. the default \"behavior\" is likely to be \"average\".. you need to very careful of what you are asking for. For me this is just a fun experiment and very interesting to see the market analysis it", "context": "", "use_case": "building stock portfolios", "switched_from": "o3", "story": "We gave 5 LLMs $100K to trade stocks for 8 months"}]}], "caveats": [{"label": "", "summary": "", "quotes": [{"id": 763237329, "item_id": 47702333, "url": "https://news.ycombinator.com/item?id=47702333", "by": "heijmans", "date": "2026-04-09", "stance": "warn", "quote": "GPT-5.4 Thinking, Pro account.", "excerpt": "Here is small example of ChatGPT giving a wrong answer without expressing any doubt (aka bullshitting): https://chatgpt.com/share/69d78ec7-67b0-8395-9fd1-522b760ab57a GPT-5.4 Thinking, Pro account.", "context": "", "use_case": "", "switched_from": "", "story": "The Future of Everything Is Lies, I Guess"}, {"id": 734212275, "item_id": 45888267, "url": "https://news.ycombinator.com/item?id=45888267", "by": "killerstorm", "date": "2025-11-11", "stance": "warn", "quote": "So, actually, no, it cannot read it, but it believes it can.", "excerpt": "Tangentially related: AI assistants (ChatGPT, Claude and Gemini) are unable to get public code from GitHub. (I.e. specifically assistants you use via web site, not Codex, Claude Code, etc.) Again, as it might be hard to believe, as situation is rather insane: flagship AI assistants cannot get publicly available code. They can get bits and pieces from README, but that might degrade response quality as it's often based", "context": "", "use_case": "", "switched_from": "", "story": "AI documentation you can talk to, for every repo"}]}], "who": [{"by": "andy12_", "context": "ML researcher", "url": "https://news.ycombinator.com/item?id=46601915", "date": "2026-01-13"}], "tags": [], "use_cases": ["searches", "ml research", "building stock portfolios", "math programming problem"], "compared": [{"slug": "chatgpt", "name": "ChatGPT", "co": 3, "switched_from": 0, "switched_to": 0}, {"slug": "o3", "name": "o3", "co": 1, "switched_from": 1, "switched_to": 0}, {"slug": "claude-opus", "name": "Claude Opus", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "claude", "name": "Claude", "co": 2, "switched_from": 0, "switched_to": 0}, {"slug": "gemini", "name": "Gemini", "co": 2, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47265045", "title": "GPT-5.4 Thinking System Card", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=47689648", "title": "The Future of Everything Is Lies, I Guess", "mentions": 2}, {"url": "https://news.ycombinator.com/item?id=47128138", "title": "\"Car Wash\" test with 53 models", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=47031580", "title": "I want to wash my car. The car wash is 50 meters away. Should I walk or drive?", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46655745", "title": "GPT-5.2 does not follow instructions and ignores my prompts", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46590280", "title": "TimeCapsuleLLM: LLM trained only on data from 1800-1875", "mentions": 1}], "sentiment_by_month": {"2025-11": {"recommend": 2, "neutral": 2, "warn": 1}, "2025-12": {"recommend": 1}, "2026-01": {"recommend": 2}, "2026-02": {"neutral": 2}, "2026-03": {"neutral": 2, "recommend": 1}, "2026-04": {"recommend": 1, "warn": 1}, "2026-05": {}, "2026-06": {}, "2026-07": {}, "2026-08": {}, "2026-09": {}, "2026-10": {}}, "quotes": [{"id": 763237329, "item_id": 47702333, "url": "https://news.ycombinator.com/item?id=47702333", "by": "heijmans", "date": "2026-04-09", "stance": "warn", "quote": "GPT-5.4 Thinking, Pro account.", "excerpt": "Here is small example of ChatGPT giving a wrong answer without expressing any doubt (aka bullshitting): https://chatgpt.com/share/69d78ec7-67b0-8395-9fd1-522b760ab57a GPT-5.4 Thinking, Pro account.", "context": "", "use_case": "", "switched_from": "", "story": "The Future of Everything Is Lies, I Guess"}, {"id": 763108432, "item_id": 47694277, "url": "https://news.ycombinator.com/item?id=47694277", "by": "simianwords", "date": "2026-04-08", "stance": "recommend", "quote": "It doesn't bullshit on the GPT-5.4 thinking version.", "excerpt": "I wanted this challenge with the thinking version (I apologised for it and edited the earlier version). It doesn't bullshit on the GPT-5.4 thinking version. Here is the result with thinking https://chatgpt.com/share/69d69dd6-fb50-838d-863c-4e1eda5d088b I suggest you try it yourself to be convinced. Try it in incognito mode if you wish. Or not.", "context": "", "use_case": "", "switched_from": "", "story": "The Future of Everything Is Lies, I Guess"}, {"id": 756362882, "item_id": 47272680, "url": "https://news.ycombinator.com/item?id=47272680", "by": "andy12_", "date": "2026-03-06", "stance": "recommend", "quote": "It concisely replies \"GPT-5.4 Thinking\".", "excerpt": "It's not a rumor; you can just test it. Ask the router \"What model are you\". It will yap on and on about being a GPT-5.3 model (Non-thinking models of OpenAI are insufferable yappers that don't know when to shut up). Ask it now \"What model are you. Think carefully\". It concisely replies \"GPT-5.4 Thinking\". https://openai.com/index/introducing-gpt-5/ > GPT\u20115 is a unified system with a smart, efficient model that answe", "context": "", "use_case": "", "switched_from": "", "story": "GPT-5.4 Thinking System Card"}, {"id": 756298032, "item_id": 47268627, "url": "https://news.ycombinator.com/item?id=47268627", "by": "wahnfrieden", "date": "2026-03-05", "stance": "neutral", "quote": "some non-employees have been using it for a while already", "excerpt": "some non-employees have been using it for a while already", "context": "", "use_case": "", "switched_from": "", "story": "GPT-5.4 Thinking System Card"}, {"id": 756245152, "item_id": 47265322, "url": "https://news.ycombinator.com/item?id=47265322", "by": "egonschiele", "date": "2026-03-05", "stance": "neutral", "quote": "The actual card is here https://deploymentsafety.openai.com/gpt-5-4-thinking/introduction", "excerpt": "The actual card is here https://deploymentsafety.openai.com/gpt-5-4-thinking/introduction the link currently goes to the announcement.", "context": "", "use_case": "", "switched_from": "", "story": "GPT-5.4 Thinking System Card"}, {"id": 754132144, "item_id": 47133259, "url": "https://news.ycombinator.com/item?id=47133259", "by": "handoflixue", "date": "2026-02-23", "stance": "neutral", "quote": "GPT 5.2 Thinking, Standard Effort: Walk", "excerpt": "Testing some subset X does not mean the test is rigged unless they failed to disclose that. But also: GPT 5.2 Thinking, Standard Effort: Walk - https://chatgpt.com/share/699d38cb-e560-8012-8986-d27428de8a01 I'm assuming \"GPT 5.2 Thinking\" is, in fact, a thinking model?", "context": "", "use_case": "", "switched_from": "", "story": "\"Car Wash\" test with 53 models"}, {"id": 752687392, "item_id": 47042962, "url": "https://news.ycombinator.com/item?id=47042962", "by": "jason_oster", "date": "2026-02-16", "stance": "neutral", "quote": "GPT-5.2 Thinking: https://chatgpt.com/share/6993d099-ef4c-8005-aa62-bdb826b707a3", "excerpt": "> Doesn't offering two options to the LLM, \"walk,\" or \"drive,\" imply that either can be chosen? Yes, but the problem is specifically that proposing two choices also eliminates other options. An open-ended question would lift that restriction. GPT-5.2 Thinking: https://chatgpt.com/share/6993d099-ef4c-8005-aa62-bdb826b707a3 Other possible issues in the question: Is biking also an option? What do you want to do at the c", "context": "", "use_case": "", "switched_from": "", "story": "I want to wash my car. The car wash is 50 meters away. Should I walk or drive?"}, {"id": 746493264, "item_id": 46655829, "url": "https://news.ycombinator.com/item?id=46655829", "by": "timpera", "date": "2026-01-16", "stance": "recommend", "quote": "I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly.", "excerpt": "People in the comments are all talking about a degradation in performance, but I used GPT-5.2-Thinking for some searches yesterday and it followed my instructions perfectly. There's no way to know for sure if the model is experiencing issues, that's the problem with this non-deterministic tech.", "context": "", "use_case": "searches", "switched_from": "", "story": "GPT-5.2 does not follow instructions and ignores my prompts"}, {"id": 745630641, "item_id": 46601915, "url": "https://news.ycombinator.com/item?id=46601915", "by": "andy12_", "date": "2026-01-13", "stance": "recommend", "quote": "When I use GPT 5.2-thinking for my ML research job, I pretty much never encounter hallucinations.", "excerpt": "> I don't entirely understand your extreme optimism towards LLMs given this proclivity for hallucination Simply because I don't see hallucinations as a permanent problem. I see that models keep improving more and more in this regard, and I don't see why the hallucination rate can't be abirtrarily reduced with further improvements to the architecture. When I ask Claude about obscure topics, it correctly replies \"I don", "context": "ML researcher", "use_case": "ML research", "switched_from": "", "story": "TimeCapsuleLLM: LLM trained only on data from 1800-1875"}, {"id": 738480465, "item_id": 46155029, "url": "https://news.ycombinator.com/item?id=46155029", "by": "XenophileJKO", "date": "2025-12-04", "stance": "recommend", "quote": "I started with o3 and now I'm using 5.1 Thinking (set to max)", "excerpt": "So.. I have been using an LLM to make 30 day buy and hold portfolios. And the results are \"ok\". (Like 8% vs 6% for the S&P 500 over the last 90 days) What you ask the model to do is super important. Just like writing or coding.. the default \"behavior\" is likely to be \"average\".. you need to very careful of what you are asking for. For me this is just a fun experiment and very interesting to see the market analysis it", "context": "", "use_case": "building stock portfolios", "switched_from": "o3", "story": "We gave 5 LLMs $100K to trade stocks for 8 months"}, {"id": 735506656, "item_id": 45969166, "url": "https://news.ycombinator.com/item?id=45969166", "by": "iamdelirium", "date": "2025-11-18", "stance": "neutral", "quote": "The 17.6% is for 5.1 Thinking High.", "excerpt": "This is provably false. All it takes is a simple Google search and looking at the ARC AGI 2 leaderboard: https://arcprize.org/leaderboard The 17.6% is for 5.1 Thinking High.", "context": "", "use_case": "", "switched_from": "", "story": "Gemini 3 Pro Model Card [pdf]"}, {"id": 734526000, "item_id": 45907875, "url": "https://news.ycombinator.com/item?id=45907875", "by": "webo", "date": "2025-11-12", "stance": "recommend", "quote": "ChatGPT-5.1-Thinking one-shotted a correct Python script without any library use", "excerpt": "I was curious if LLMs are good for this problem. ChatGPT-5.1-Thinking one-shotted a correct Python script without any library use (https://pastecode.io/s/jg6ggxpm). Claude Opus failed to solve after trying for a while.", "context": "", "use_case": "Math programming problem", "switched_from": "", "story": "Project Euler"}, {"id": 734212275, "item_id": 45888267, "url": "https://news.ycombinator.com/item?id=45888267", "by": "killerstorm", "date": "2025-11-11", "stance": "warn", "quote": "So, actually, no, it cannot read it, but it believes it can.", "excerpt": "Tangentially related: AI assistants (ChatGPT, Claude and Gemini) are unable to get public code from GitHub. (I.e. specifically assistants you use via web site, not Codex, Claude Code, etc.) Again, as it might be hard to believe, as situation is rather insane: flagship AI assistants cannot get publicly available code. They can get bits and pieces from README, but that might degrade response quality as it's often based", "context": "", "use_case": "", "switched_from": "", "story": "AI documentation you can talk to, for every repo"}, {"id": 733822257, "item_id": 45863891, "url": "https://news.ycombinator.com/item?id=45863891", "by": "danskeren", "date": "2025-11-09", "stance": "neutral", "quote": "I throw all my queries at Grok 4 Expert, GPT 5 Thinking and Opus 4.1 Extended Thinking", "excerpt": "I throw all my queries at Grok 4 Expert, GPT 5 Thinking and Opus 4.1 Extended Thinking.. for Golang it's been my experience that Grok produce the best results about 90% of the time as well. Some simple example: https://claude.ai/share/6d178173-cdf7-4e50-a467-73ee9f479d56. https://chatgpt.com/share/69102735-46ac-8012-9cf0-0969585c86e6. https://grok.com/share/bGVnYWN5LWNvcHk%3D_54b5f2f1-732e-4372-b0bb-d72f4ac31004. I d", "context": "", "use_case": "", "switched_from": "", "story": null}, {"id": 732766626, "item_id": 45797914, "url": "https://news.ycombinator.com/item?id=45797914", "by": "simonw", "date": "2025-11-03", "stance": "recommend", "quote": "almost always opts to run a search before it attempts to answer a question", "excerpt": "Claude Code has a neat fix for that - it knows to look at its own documentation if you ask it questions about itself: https://simonwillison.net/2025/Oct/24/claude-code-docs-map/ I've had great results from ChatGPT running the \"GPT-5 Thinking\" model since that almost always opts to run a search before it attempts to answer a question. Here's what I got from that for your Switch 2 question: https://chatgpt.com/share/69", "context": "", "use_case": "", "switched_from": "", "story": "Claude Code can debug low-level cryptography"}]}