{"slug": "cursor-composer", "name": "Cursor Composer", "category": "ai-models", "category_name": "AI models", "description": "Cursor's in-house coding model for agentic tasks inside the Cursor AI code editor.", "recommenders": 41, "commenters": 65, "warners": 12, "score": 29.079, "rising": -6.78, "first": "2025-11-15", "latest": "2026-09-26", "why": "HN commenters use it as a fast, cheap daily workhorse for small and routine coding tasks, often handing it plans written by stronger models.", "why_is_quote": false, "has_verdict": true, "sentiment": {"recommend": 41, "neutral": 10, "compare": 4, "warn": 12}, "spark": [3, 0, 0, 0, 0, 5, 3, 14, 9, 3, 4, 0], "listed": true, "loved": ["Speed", "Low cost"], "caveat": "Complex tasks", "short_quote": "I find it really hard to justify the other models for the performance/cost I'm getting with Composer 2.5.", "kind": "", "icon": true, "domain": "cursor.com", "github_repo": null, "official_url": "https://cursor.com", "pricing": "freemium", "aliases": ["Composer 2.5", "Composer", "Composer 2", "Cursor Composer 2", "Composer 1", "composer-2", "Composer-1"], "verdict": [{"text": "HN commenters use it as a fast, cheap daily workhorse", "cites": [{"id": 794776992, "item_id": 49673562, "url": "https://news.ycombinator.com/item?id=49673562", "by": "time0ut", "date": "2026-09-12", "version": null, "stance": "recommend", "quote": "It is just a cheap, fast, surgical workhorse.", "excerpt": "If I had to pick just one, these days it\u2019s Composer. It is just a cheap, fast, surgical workhorse. My workflow uses other models as well for various phases: Opus for planning, Sonnet for review, etc. I prefer Cursor at this point just because of Composer. Claude Code is passable but the lack of a good, fast, cheap workhorse sucks. Sonnet and Haiku aren\u2019t it. I also do like Codex and Sol, Terra, and Luna. They are dec", "context": "", "use_case": "implementation", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 780117184, "item_id": 48757324, "url": "https://news.ycombinator.com/item?id=48757324", "by": "danfritz", "date": "2026-07-01", "version": "2.5", "stance": "recommend", "quote": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done.", "excerpt": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done. I only reach for Claud when i need to plan something big or want to have a sparring partner to fire of some ideas. I think what a lot of people don't realize is that you don't need a fronteer model for 80% of coding tasks. Composer 2.5 is often more than good enough, less token hungry and way faster", "context": "", "use_case": "daily coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780430739, "item_id": 48776921, "url": "https://news.ycombinator.com/item?id=48776921", "by": "jarodrh", "date": "2026-07-03", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 (brilliant and fast)", "excerpt": "Orchestration works very well for me, but not in the way most people seem to be pushing for, with middlemen scoring and routing every request. I'm not sure I agree with that direction. For coding agents the routing problem is mostly solved at the config level, the harness already lets you pin models per role (done via subagents), and that covers most of what a per-request router promises. Three things have happened t", "context": "", "use_case": "agent pool coding", "switched_from": "", "story": "Ask HN: Is anyone experimenting with different ways of using LLMs for coding?"}]}, {"text": " for small and routine coding tasks,", "cites": [{"id": 797680064, "item_id": 49855004, "url": "https://news.ycombinator.com/item?id=49855004", "by": "the_other", "date": "2026-09-26", "version": "2.5", "stance": "recommend", "quote": "I do the bulk of my work with Composer 2.5. It\u2019s quite good for \u201cgoing forward\u201d on smallish tasks", "excerpt": "The last couple of weeks, I\u2019ve got rigorous about making the agent refactor old code. We\u2019ve removed, moved, consolidated, reordered loads of cruft. It has made the code much tidier and reduced the chance that the next feature will build on top of cruft. E.G. over time we\u2019d gained two client-side caches of related server state. This started out as two different parts of the same model, because we couldn\u2019t get all the ", "context": "employer provides limited LLM budget", "use_case": "daily coding tasks", "switched_from": "", "story": "Plan mode is dead"}, {"id": 777034048, "item_id": 48564628, "url": "https://news.ycombinator.com/item?id=48564628", "by": "ghshephard", "date": "2026-06-16", "version": null, "stance": "recommend", "quote": "that I use a lot of the time when I have fairly straightforward tasks", "excerpt": "You nailed it - in fact, most of Anthropic's early revenue came from Cursor - much of claude code programming components is essentially a feature copy of Cursor, so it makes sense they are similar. Cursor does have it's own model - it's a heavily reworked version of KimiK2, called \"composer\" - that I use a lot of the time when I have fairly straightforward tasks that don't require a lot of exploration or independent ", "context": "", "use_case": "straightforward coding tasks", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 780262112, "item_id": 48766382, "url": "https://news.ycombinator.com/item?id=48766382", "by": "soyin", "date": "2026-07-02", "version": "2.5", "stance": "recommend", "quote": "I find it really hard to justify the other models for the performance/cost I'm getting with Composer 2.5.", "excerpt": "I'm also using it as my daily driver. I've been trying Opus 4.8 this week to see if I was missing something but haven't noticed a meaningful difference. I'm working on a fairly routine full stack web app that isn't doing anything incredible. Once I had the patterns I wanted in place, it's been very capable of following those with new work. I also don't ever give it long running tasks, it's always focused and small ch", "context": "working on a fairly routine full stack web app", "use_case": "full stack web app coding", "switched_from": "", "story": "CursorBench 3.1"}]}, {"text": " often handing it plans written by stronger models.", "cites": [{"id": 796707840, "item_id": 49794240, "url": "https://news.ycombinator.com/item?id=49794240", "by": "the_sleaze_", "date": "2026-09-21", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 - which remains incredible and honestly nothing else compares for me", "excerpt": "I didn't find 4.6 any better than Composer 2.5 - which remains incredible and honestly nothing else compares for me. Make a galaxy model search the space, create a document, argue and defend decisions, then hand it to 2.5 to implement. 4.6 was a slower less enjoyable version of that. 4.7 is better at \"I want the button to cancel the jobs, dont make any mistakes\" but honestly that's not what I use it's class for.", "context": "", "use_case": "implementing plans from documents", "switched_from": "", "story": "Grok 4.7"}, {"id": 791794673, "item_id": 49487167, "url": "https://news.ycombinator.com/item?id=49487167", "by": "rippeltippel", "date": "2026-08-29", "version": null, "stance": "recommend", "quote": "their free tier of Composer is good enough", "excerpt": "I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time. Using both OpenAI and Anthropic models, Cursor is cheaper than", "context": "", "use_case": "implementation", "switched_from": "", "story": "Our decision on Cursor following its acquisition by SpaceX"}, {"id": 794776992, "item_id": 49673562, "url": "https://news.ycombinator.com/item?id=49673562", "by": "time0ut", "date": "2026-09-12", "version": null, "stance": "recommend", "quote": "It is just a cheap, fast, surgical workhorse.", "excerpt": "If I had to pick just one, these days it\u2019s Composer. It is just a cheap, fast, surgical workhorse. My workflow uses other models as well for various phases: Opus for planning, Sonnet for review, etc. I prefer Cursor at this point just because of Composer. Claude Code is passable but the lack of a good, fast, cheap workhorse sucks. Sonnet and Haiku aren\u2019t it. I also do like Codex and Sol, Terra, and Luna. They are dec", "context": "", "use_case": "implementation", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}]}, {"text": " The most common complaint is shallow or mistake-prone results on more complex work,", "cites": [{"id": 780313664, "item_id": 48769604, "url": "https://news.ycombinator.com/item?id=48769604", "by": "Losenok", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow.", "excerpt": "Very skeptical about the composer accuracy. I have been using it for 6 months now and it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow. It feels like it just finds the cheapest way to deliver the task.", "context": "using it 6 months", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 781370464, "item_id": 48835654, "url": "https://news.ycombinator.com/item?id=48835654", "by": "bhouston", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes.", "excerpt": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes. It is unusable compared to GPT 5.4 or 5.5. My time is more valuable that I will use a model that doesn\u2019t f** up my code base.", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Grok 4.5"}, {"id": 780997952, "item_id": 48812372, "url": "https://news.ycombinator.com/item?id=48812372", "by": "richkuo", "date": "2026-07-06", "version": "2.5", "stance": "recommend", "quote": "Incredibly token efficient, fast, and gets the same quality as if you worked the whole thing with Fable/Opus/GPT.", "excerpt": "Composer 2.5 is really effective at some tasks, but doesn't do as well on higher complexity tasks from my own experience. With that being said, Cursor Bench is nonetheless a fantastic gauge of LLM quality and performance. The most interesting outcome of Cursor Bench is Fable 5 to GPT 5.5, its almost a perfect continuous line. As of the report, Fable to GPT might be the new standard of agentic programming/building I d", "context": "open source maintainer", "use_case": "building code after planning", "switched_from": "", "story": "CursorBench 3.1"}]}, {"text": " where several prefer Opus or GPT models.", "cites": [{"id": 780265537, "item_id": 48766596, "url": "https://news.ycombinator.com/item?id=48766596", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "Composer hands me a quickly put together college project that was started the night before it was due.", "excerpt": "I'm not disputing what you're saying. The slowness of Opus in particular is pretty accurate but you should have been getting little popups in Cursor saying Opus is under load and to try switching to other models? I mentioned my frustration with Composer in another thread and why I rely on Opus, but Opus, atleast via Cursor is practically unusable for me M-F 9-5 EST. As a result, I have modified my working schedule ou", "context": "", "use_case": "AI coding via Cursor", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780269488, "item_id": 48766843, "url": "https://news.ycombinator.com/item?id=48766843", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "my distaste for Composer and my preference for Opus", "excerpt": "> it tends to leave big, dangerous holes hiding inside implementations unless babied it's fascinating that I used these same exact words to express my distaste for Composer and my preference for Opus. I suspect, the domains and problems we are trying to solve need to be shared. I wrote about it here: https://news.ycombinator.com/item?id=48766275 Would love to reach out to discuss more, if you're ok with it, or absolu", "context": "", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 781370464, "item_id": 48835654, "url": "https://news.ycombinator.com/item?id=48835654", "by": "bhouston", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes.", "excerpt": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes. It is unusable compared to GPT 5.4 or 5.5. My time is more valuable that I will use a model that doesn\u2019t f** up my code base.", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Grok 4.5"}]}], "praise": [{"label": "Speed", "summary": "Many call it very fast, including compared with Anthropic models.", "quotes": [{"id": 777084561, "item_id": 48567785, "url": "https://news.ycombinator.com/item?id=48567785", "by": "jemmyw", "date": "2026-06-17", "version": null, "stance": "recommend", "quote": "Composer is actually really good, and fast.", "excerpt": "I didn't like cursor when it first came about but now I use it for my personal projects. The plan is good value for accessing different sota models occasionally for planning. Composer is actually really good, and fast. I'm not sure if it's because my personal projects are small enough I know them inside out and my work project is huge, but I prefer terminal code agents for day work over ide integration.", "context": "", "use_case": "", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 776893616, "item_id": 48555851, "url": "https://news.ycombinator.com/item?id=48555851", "by": "infecto", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "I use it quite a bit and it works well and is extremely fast for the tasks it was built for", "excerpt": "Hmmm. Not in my experience. I don\u2019t think it can be compared to Haiku, maybe sonnet levels? It\u2019s obviously not Opus and never was intended for that use case. I use it quite a bit and it works well and is extremely fast for the tasks it was built for", "context": "", "use_case": "fast coding tasks", "switched_from": "", "story": "SpaceX Is Buying Cursor"}, {"id": 780313664, "item_id": 48769604, "url": "https://news.ycombinator.com/item?id=48769604", "by": "Losenok", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow.", "excerpt": "Very skeptical about the composer accuracy. I have been using it for 6 months now and it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow. It feels like it just finds the cheapest way to deliver the task.", "context": "using it 6 months", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}]}, {"label": "Low cost", "summary": "Commenters say it is cheap enough for daily use under budgets, with a usable free tier.", "quotes": [{"id": 794776992, "item_id": 49673562, "url": "https://news.ycombinator.com/item?id=49673562", "by": "time0ut", "date": "2026-09-12", "version": null, "stance": "recommend", "quote": "It is just a cheap, fast, surgical workhorse.", "excerpt": "If I had to pick just one, these days it\u2019s Composer. It is just a cheap, fast, surgical workhorse. My workflow uses other models as well for various phases: Opus for planning, Sonnet for review, etc. I prefer Cursor at this point just because of Composer. Claude Code is passable but the lack of a good, fast, cheap workhorse sucks. Sonnet and Haiku aren\u2019t it. I also do like Codex and Sol, Terra, and Luna. They are dec", "context": "", "use_case": "implementation", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 777034048, "item_id": 48564628, "url": "https://news.ycombinator.com/item?id=48564628", "by": "ghshephard", "date": "2026-06-16", "version": null, "stance": "recommend", "quote": "that I use a lot of the time when I have fairly straightforward tasks", "excerpt": "You nailed it - in fact, most of Anthropic's early revenue came from Cursor - much of claude code programming components is essentially a feature copy of Cursor, so it makes sense they are similar. Cursor does have it's own model - it's a heavily reworked version of KimiK2, called \"composer\" - that I use a lot of the time when I have fairly straightforward tasks that don't require a lot of exploration or independent ", "context": "", "use_case": "straightforward coding tasks", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 780262112, "item_id": 48766382, "url": "https://news.ycombinator.com/item?id=48766382", "by": "soyin", "date": "2026-07-02", "version": "2.5", "stance": "recommend", "quote": "I find it really hard to justify the other models for the performance/cost I'm getting with Composer 2.5.", "excerpt": "I'm also using it as my daily driver. I've been trying Opus 4.8 this week to see if I was missing something but haven't noticed a meaningful difference. I'm working on a fairly routine full stack web app that isn't doing anything incredible. Once I had the patterns I wanted in place, it's been very capable of following those with new work. I also don't ever give it long running tasks, it's always focused and small ch", "context": "working on a fairly routine full stack web app", "use_case": "full stack web app coding", "switched_from": "", "story": "CursorBench 3.1"}]}, {"label": "Routine tasks", "summary": "Described as a dependable daily driver for small, focused, straightforward work.", "quotes": [{"id": 797680064, "item_id": 49855004, "url": "https://news.ycombinator.com/item?id=49855004", "by": "the_other", "date": "2026-09-26", "version": "2.5", "stance": "recommend", "quote": "I do the bulk of my work with Composer 2.5. It\u2019s quite good for \u201cgoing forward\u201d on smallish tasks", "excerpt": "The last couple of weeks, I\u2019ve got rigorous about making the agent refactor old code. We\u2019ve removed, moved, consolidated, reordered loads of cruft. It has made the code much tidier and reduced the chance that the next feature will build on top of cruft. E.G. over time we\u2019d gained two client-side caches of related server state. This started out as two different parts of the same model, because we couldn\u2019t get all the ", "context": "employer provides limited LLM budget", "use_case": "daily coding tasks", "switched_from": "", "story": "Plan mode is dead"}, {"id": 780117184, "item_id": 48757324, "url": "https://news.ycombinator.com/item?id=48757324", "by": "danfritz", "date": "2026-07-01", "version": "2.5", "stance": "recommend", "quote": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done.", "excerpt": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done. I only reach for Claud when i need to plan something big or want to have a sparring partner to fire of some ideas. I think what a lot of people don't realize is that you don't need a fronteer model for 80% of coding tasks. Composer 2.5 is often more than good enough, less token hungry and way faster", "context": "", "use_case": "daily coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 777034048, "item_id": 48564628, "url": "https://news.ycombinator.com/item?id=48564628", "by": "ghshephard", "date": "2026-06-16", "version": null, "stance": "recommend", "quote": "that I use a lot of the time when I have fairly straightforward tasks", "excerpt": "You nailed it - in fact, most of Anthropic's early revenue came from Cursor - much of claude code programming components is essentially a feature copy of Cursor, so it makes sense they are similar. Cursor does have it's own model - it's a heavily reworked version of KimiK2, called \"composer\" - that I use a lot of the time when I have fairly straightforward tasks that don't require a lot of exploration or independent ", "context": "", "use_case": "straightforward coding tasks", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}]}, {"label": "Implementation partner", "summary": "Used to implement plans written by larger models such as Opus.", "quotes": [{"id": 796707840, "item_id": 49794240, "url": "https://news.ycombinator.com/item?id=49794240", "by": "the_sleaze_", "date": "2026-09-21", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 - which remains incredible and honestly nothing else compares for me", "excerpt": "I didn't find 4.6 any better than Composer 2.5 - which remains incredible and honestly nothing else compares for me. Make a galaxy model search the space, create a document, argue and defend decisions, then hand it to 2.5 to implement. 4.6 was a slower less enjoyable version of that. 4.7 is better at \"I want the button to cancel the jobs, dont make any mistakes\" but honestly that's not what I use it's class for.", "context": "", "use_case": "implementing plans from documents", "switched_from": "", "story": "Grok 4.7"}, {"id": 794776992, "item_id": 49673562, "url": "https://news.ycombinator.com/item?id=49673562", "by": "time0ut", "date": "2026-09-12", "version": null, "stance": "recommend", "quote": "It is just a cheap, fast, surgical workhorse.", "excerpt": "If I had to pick just one, these days it\u2019s Composer. It is just a cheap, fast, surgical workhorse. My workflow uses other models as well for various phases: Opus for planning, Sonnet for review, etc. I prefer Cursor at this point just because of Composer. Claude Code is passable but the lack of a good, fast, cheap workhorse sucks. Sonnet and Haiku aren\u2019t it. I also do like Codex and Sol, Terra, and Luna. They are dec", "context": "", "use_case": "implementation", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 791794673, "item_id": 49487167, "url": "https://news.ycombinator.com/item?id=49487167", "by": "rippeltippel", "date": "2026-08-29", "version": null, "stance": "recommend", "quote": "their free tier of Composer is good enough", "excerpt": "I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time. Using both OpenAI and Anthropic models, Cursor is cheaper than", "context": "", "use_case": "implementation", "switched_from": "", "story": "Our decision on Cursor following its acquisition by SpaceX"}]}, {"label": "Token efficiency", "summary": "Some say it uses fewer tokens while matching larger models on some tasks.", "quotes": [{"id": 780997952, "item_id": 48812372, "url": "https://news.ycombinator.com/item?id=48812372", "by": "richkuo", "date": "2026-07-06", "version": "2.5", "stance": "recommend", "quote": "Incredibly token efficient, fast, and gets the same quality as if you worked the whole thing with Fable/Opus/GPT.", "excerpt": "Composer 2.5 is really effective at some tasks, but doesn't do as well on higher complexity tasks from my own experience. With that being said, Cursor Bench is nonetheless a fantastic gauge of LLM quality and performance. The most interesting outcome of Cursor Bench is Fable 5 to GPT 5.5, its almost a perfect continuous line. As of the report, Fable to GPT might be the new standard of agentic programming/building I d", "context": "open source maintainer", "use_case": "building code after planning", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 783782241, "item_id": 48986390, "url": "https://news.ycombinator.com/item?id=48986390", "by": "striking", "date": "2026-07-20", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 is itself a fine-tune of K2.7, and it's apparently quite token efficient", "excerpt": "Sure, let's have a look... > I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence. [emphasis mine] I guess I'm missing the part of this article where they bring hard numbers in to back up the argum", "context": "", "use_case": "", "switched_from": "", "story": "Who's Afraid of Chinese Models?"}, {"id": 780117184, "item_id": 48757324, "url": "https://news.ycombinator.com/item?id=48757324", "by": "danfritz", "date": "2026-07-01", "version": "2.5", "stance": "recommend", "quote": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done.", "excerpt": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done. I only reach for Claud when i need to plan something big or want to have a sparring partner to fire of some ideas. I think what a lot of people don't realize is that you don't need a fronteer model for 80% of coding tasks. Composer 2.5 is often more than good enough, less token hungry and way faster", "context": "", "use_case": "daily coding", "switched_from": "", "story": "CursorBench 3.1"}]}, {"label": "Code search", "summary": "One user finds it fast for code search, explanation and Q&A.", "quotes": [{"id": 776963841, "item_id": 48560240, "url": "https://news.ycombinator.com/item?id=48560240", "by": "darklinear", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 is really fast in their harness at code search/explanation/Q&A tasks", "excerpt": "It's still my primary at work (data engineering/platform engineering), running a mix of GPT-5.5 and Composer 2.5. We also have Claude Code subscriptions. I find myself preferring Cursor for most tasks. 70% of the time I use AI agents on a pretty tight leash. I often reject edits and ask it to change things. The IDE integration is really efficient for this workflow compared to Claude (yes, I've tried the Claude extens", "context": "data engineering/platform engineering", "use_case": "code search and debugging", "switched_from": "", "story": "SpaceX Is Buying Cursor"}]}], "caveats": [{"label": "Complex tasks", "summary": "Users report shallow or weaker results on more difficult tasks.", "quotes": [{"id": 780313664, "item_id": 48769604, "url": "https://news.ycombinator.com/item?id=48769604", "by": "Losenok", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow.", "excerpt": "Very skeptical about the composer accuracy. I have been using it for 6 months now and it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow. It feels like it just finds the cheapest way to deliver the task.", "context": "using it 6 months", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780997952, "item_id": 48812372, "url": "https://news.ycombinator.com/item?id=48812372", "by": "richkuo", "date": "2026-07-06", "version": "2.5", "stance": "recommend", "quote": "Incredibly token efficient, fast, and gets the same quality as if you worked the whole thing with Fable/Opus/GPT.", "excerpt": "Composer 2.5 is really effective at some tasks, but doesn't do as well on higher complexity tasks from my own experience. With that being said, Cursor Bench is nonetheless a fantastic gauge of LLM quality and performance. The most interesting outcome of Cursor Bench is Fable 5 to GPT 5.5, its almost a perfect continuous line. As of the report, Fable to GPT might be the new standard of agentic programming/building I d", "context": "open source maintainer", "use_case": "building code after planning", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780265537, "item_id": 48766596, "url": "https://news.ycombinator.com/item?id=48766596", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "Composer hands me a quickly put together college project that was started the night before it was due.", "excerpt": "I'm not disputing what you're saying. The slowness of Opus in particular is pretty accurate but you should have been getting little popups in Cursor saying Opus is under load and to try switching to other models? I mentioned my frustration with Composer in another thread and why I rely on Opus, but Opus, atleast via Cursor is practically unusable for me M-F 9-5 EST. As a result, I have modified my working schedule ou", "context": "", "use_case": "AI coding via Cursor", "switched_from": "", "story": "CursorBench 3.1"}]}, {"label": "Mistakes", "summary": "Some say they spend time cleaning up its errors and rate it below GPT models.", "quotes": [{"id": 781370464, "item_id": 48835654, "url": "https://news.ycombinator.com/item?id=48835654", "by": "bhouston", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes.", "excerpt": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes. It is unusable compared to GPT 5.4 or 5.5. My time is more valuable that I will use a model that doesn\u2019t f** up my code base.", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Grok 4.5"}, {"id": 781395712, "item_id": 48837232, "url": "https://news.ycombinator.com/item?id=48837232", "by": "qingcharles", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Composer 2.5 is worse than both; I use it all day for simple stuff.", "excerpt": "Composer 2.5 is worse than both; I use it all day for simple stuff. It's Kimi 2.6 in a new outfit.", "context": "", "use_case": "simple coding tasks", "switched_from": "", "story": "SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence"}]}, {"label": "Needs supervision", "summary": "Commenters say it needs guidance or can leave hidden problems unless watched.", "quotes": [{"id": 780269488, "item_id": 48766843, "url": "https://news.ycombinator.com/item?id=48766843", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "my distaste for Composer and my preference for Opus", "excerpt": "> it tends to leave big, dangerous holes hiding inside implementations unless babied it's fascinating that I used these same exact words to express my distaste for Composer and my preference for Opus. I suspect, the domains and problems we are trying to solve need to be shared. I wrote about it here: https://news.ycombinator.com/item?id=48766275 Would love to reach out to discuss more, if you're ok with it, or absolu", "context": "", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780117184, "item_id": 48757324, "url": "https://news.ycombinator.com/item?id=48757324", "by": "danfritz", "date": "2026-07-01", "version": "2.5", "stance": "recommend", "quote": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done.", "excerpt": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done. I only reach for Claud when i need to plan something big or want to have a sparring partner to fire of some ideas. I think what a lot of people don't realize is that you don't need a fronteer model for 80% of coding tasks. Composer 2.5 is often more than good enough, less token hungry and way faster", "context": "", "use_case": "daily coding", "switched_from": "", "story": "CursorBench 3.1"}]}, {"label": "Benchmark doubts", "summary": "One commenter cites third-party benchmarks ranking it far below Cursor's own claims.", "quotes": [{"id": 780119777, "item_id": 48757486, "url": "https://news.ycombinator.com/item?id=48757486", "by": "mdasen", "date": "2026-07-01", "version": "2.5", "stance": "warn", "quote": "Artificial Analysis' testing shows Composer 2.5 to be pretty far behind", "excerpt": "I'm a bit skeptical. Cursor's benchmark finds that Cursor's model (Composer 2.5) is basically as good as Opus 4.8 max and GPT-5.5 xhigh, but at a fraction of the price. Artificial Analysis' testing shows Composer 2.5 to be pretty far behind: https://artificialanalysis.ai/agents/coding-agents. You look at the DeepSWE benchmark (which is probably the hardest to game at this point) and GPT-5.5 xhigh gets a 64, Opus 4.8 ", "context": "", "use_case": "", "switched_from": "", "story": "CursorBench 3.1"}]}], "who": [{"by": "the_other", "context": "employer provides limited LLM budget", "url": "https://news.ycombinator.com/item?id=49855004", "date": "2026-09-26"}, {"by": "richkuo", "context": "open source maintainer", "url": "https://news.ycombinator.com/item?id=48812372", "date": "2026-07-06"}, {"by": "soyin", "context": "working on a fairly routine full stack web app", "url": "https://news.ycombinator.com/item?id=48766382", "date": "2026-07-02"}, {"by": "ifwinterco", "context": "works under a monthly spend cap at work", "url": "https://news.ycombinator.com/item?id=48739025", "date": "2026-06-30"}, {"by": "tomerbd", "context": "codes daily with AI", "url": "https://news.ycombinator.com/item?id=48580530", "date": "2026-06-17"}, {"by": "darklinear", "context": "data engineering/platform engineering", "url": "https://news.ycombinator.com/item?id=48560240", "date": "2026-06-16"}, {"by": "senordevnyc", "context": "runs a SaaS full-time", "url": "https://news.ycombinator.com/item?id=48554254", "date": "2026-06-16"}, {"by": "wahnfrieden", "context": "solo indie developer", "url": "https://news.ycombinator.com/item?id=46075688", "date": "2025-11-27"}], "tags": ["for Cursor users", "for routine coding tasks", "for tight LLM budgets", "for implementing plans"], "use_cases": ["ai coding", "coding", "implementation", "ai coding model", "coding model", "daily coding tasks"], "compared": [{"slug": "cursor", "name": "Cursor", "co": 31, "switched_from": 0, "switched_to": 0}, {"slug": "claude-opus", "name": "Claude Opus", "co": 22, "switched_from": 0, "switched_to": 0}, {"slug": "claude-code", "name": "Claude Code", "co": 16, "switched_from": 0, "switched_to": 0}, {"slug": "kimi", "name": "Kimi", "co": 15, "switched_from": 0, "switched_to": 0}, {"slug": "gpt-5", "name": "GPT-5", "co": 14, "switched_from": 0, "switched_to": 0}, {"slug": "codex", "name": "Codex", "co": 12, "switched_from": 0, "switched_to": 0}, {"slug": "grok", "name": "Grok", "co": 10, "switched_from": 0, "switched_to": 0}, {"slug": "claude-sonnet", "name": "Claude Sonnet", "co": 10, "switched_from": 0, "switched_to": 0}], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=48756840", "title": "CursorBench 3.1", "mentions": 8}, {"url": "https://news.ycombinator.com/item?id=48553224", "title": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=48182516", "title": "Cursor Introduces Composer 2.5", "mentions": 6}, {"url": "https://news.ycombinator.com/item?id=48502347", "title": "Kimi K2.7-Code: open-source coding model with better token efficiency", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=47855293", "title": "SpaceX and Cursor partnership. Right to acquire Cursor later this year", "mentions": 3}, {"url": "https://news.ycombinator.com/item?id=49486172", "title": "Our decision on Cursor following its acquisition by SpaceX", "mentions": 2}], "sentiment_by_month": {"2025-11": {"recommend": 3}, "2025-12": {"warn": 1}, "2026-01": {}, "2026-02": {"compare": 1}, "2026-03": {"warn": 3, "neutral": 1}, "2026-04": {"recommend": 5, "neutral": 1}, "2026-05": {"neutral": 2, "warn": 3, "recommend": 3, "compare": 1}, "2026-06": {"recommend": 14, "neutral": 3, "compare": 1}, "2026-07": {"recommend": 9, "warn": 5, "compare": 1, "neutral": 1}, "2026-08": {"recommend": 3, "neutral": 2}, "2026-09": {"recommend": 4}, "2026-10": {}}, "versions": [{"v": "2.5", "people": 35, "rec": 23, "warn": 3, "other": 9, "first": "2026-05-19"}, {"v": "2", "people": 12, "rec": 6, "warn": 4, "other": 2, "first": "2026-03-20"}], "version_unnamed": 16, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 797680064, "item_id": 49855004, "url": "https://news.ycombinator.com/item?id=49855004", "by": "the_other", "date": "2026-09-26", "version": "2.5", "stance": "recommend", "quote": "I do the bulk of my work with Composer 2.5. It\u2019s quite good for \u201cgoing forward\u201d on smallish tasks", "excerpt": "The last couple of weeks, I\u2019ve got rigorous about making the agent refactor old code. We\u2019ve removed, moved, consolidated, reordered loads of cruft. It has made the code much tidier and reduced the chance that the next feature will build on top of cruft. E.G. over time we\u2019d gained two client-side caches of related server state. This started out as two different parts of the same model, because we couldn\u2019t get all the ", "context": "employer provides limited LLM budget", "use_case": "daily coding tasks", "switched_from": "", "story": "Plan mode is dead"}, {"id": 796707840, "item_id": 49794240, "url": "https://news.ycombinator.com/item?id=49794240", "by": "the_sleaze_", "date": "2026-09-21", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 - which remains incredible and honestly nothing else compares for me", "excerpt": "I didn't find 4.6 any better than Composer 2.5 - which remains incredible and honestly nothing else compares for me. Make a galaxy model search the space, create a document, argue and defend decisions, then hand it to 2.5 to implement. 4.6 was a slower less enjoyable version of that. 4.7 is better at \"I want the button to cancel the jobs, dont make any mistakes\" but honestly that's not what I use it's class for.", "context": "", "use_case": "implementing plans from documents", "switched_from": "", "story": "Grok 4.7"}, {"id": 794776992, "item_id": 49673562, "url": "https://news.ycombinator.com/item?id=49673562", "by": "time0ut", "date": "2026-09-12", "version": null, "stance": "recommend", "quote": "It is just a cheap, fast, surgical workhorse.", "excerpt": "If I had to pick just one, these days it\u2019s Composer. It is just a cheap, fast, surgical workhorse. My workflow uses other models as well for various phases: Opus for planning, Sonnet for review, etc. I prefer Cursor at this point just because of Composer. Claude Code is passable but the lack of a good, fast, cheap workhorse sucks. Sonnet and Haiku aren\u2019t it. I also do like Codex and Sol, Terra, and Luna. They are dec", "context": "", "use_case": "implementation", "switched_from": "", "story": "Ask HN: What default model do you use and why?"}, {"id": 792880017, "item_id": 49555001, "url": "https://news.ycombinator.com/item?id=49555001", "by": "lacoolj", "date": "2026-09-03", "version": "2.5", "stance": "recommend", "quote": "can be handled with Auto mode or Composer 2.5 (or Grok 4.6 High)", "excerpt": "Jesus what are you doing that requires Sol usage so often? Terra not enough? I know Luna isn't reliable, so that's fair. Genuinely curious though, because I use Cursor daily and almost everything I do, highly complex or high volume, can be handled with Auto mode or Composer 2.5 (or Grok 4.6 High). So I have to assume you're doing something far more complex than what I am", "context": "", "use_case": "coding", "switched_from": "", "story": "OpenAI begins rolling out GPT-6 Astra"}, {"id": 791913572, "item_id": 49494598, "url": "https://news.ycombinator.com/item?id=49494598", "by": "thefourthchime", "date": "2026-08-30", "version": "2.5", "stance": "neutral", "quote": "auto picks between Opus 5, GPT 5.6, Grok 4.6, and Composer 2.5", "excerpt": "It's a little confusing because Grok and Cursor kinda live in different worlds. Grok has SuperGrok and SuperGrok Heavy ($300 a month!) The best bet is Cursor Ultra ($200), which is the equivalent a month plan that you get from OpenAI and Anthropic. I get a LOT more capacity out of it because I almost always use the \"Auto\" mode, which auto picks between Opus 5, GPT 5.6, Grok 4.6, and Composer 2.5. I like Auto mode so ", "context": "", "use_case": "coding via Cursor Auto", "switched_from": "", "story": "Our decision on Cursor following its acquisition by SpaceX"}, {"id": 791794673, "item_id": 49487167, "url": "https://news.ycombinator.com/item?id=49487167", "by": "rippeltippel", "date": "2026-08-29", "version": null, "stance": "recommend", "quote": "their free tier of Composer is good enough", "excerpt": "I may be a rare Cursor user around here, but I am saddened by this news. The thing that keeps me using it is precisely the ability to switch between models: I usually do brainstorm/planning with Sol or Terra, then let Composer (free) deal with the implementation. For certain tasks I use Claude, which tends to be more expensive but sometimes saves me time. Using both OpenAI and Anthropic models, Cursor is cheaper than", "context": "", "use_case": "implementation", "switched_from": "", "story": "Our decision on Cursor following its acquisition by SpaceX"}, {"id": 790115637, "item_id": 49382227, "url": "https://news.ycombinator.com/item?id=49382227", "by": "nikolay", "date": "2026-08-20", "version": "2.5", "stance": "neutral", "quote": "Why would I need Composer unless I'm a poor hobbyist?", "excerpt": "I use Claude with the Codex plugin as a critic (sometimes I switch and manually review with Claude). With the Superpowers skill (which I am trying to get rid of), it designs expensive models and implementations, uses cheaper ones for other, less important stuff, including Haiku, sometimes upgrading if it hits a roadblock, and then, finally, reviews with expensive ones. Why would I need Composer unless I'm a poor hobb", "context": "", "use_case": "", "switched_from": "", "story": "Git at Any Scale"}, {"id": 786240578, "item_id": 49140036, "url": "https://news.ycombinator.com/item?id=49140036", "by": "eleventen", "date": "2026-08-01", "version": "2.5", "stance": "recommend", "quote": "happily using Cursor + Composer 2.5/Opus", "excerpt": "I'm surprised that their biggest feature has not been mentioned: The permissions model is much less noisy in Cursor. Every 30 seconds in claude you get a new, elaborate, contrived bash command that claude wants to run. It's some combination of 10 things you've already granted permission for, but this one is new! Cursor's auto-review run mode fixes that https://cursor.com/docs/agent/security/run-modes I know it's a co", "context": "", "use_case": "", "switched_from": "", "story": "Cursor removed cost information from the usage page and CSV export"}, {"id": 786180162, "item_id": 49136260, "url": "https://news.ycombinator.com/item?id=49136260", "by": "DoesntMatter22", "date": "2026-08-01", "version": "2.5", "stance": "recommend", "quote": "Built an awesome 450k line app with composer 2.5 and grok 4.5. I love it", "excerpt": "Built an awesome 450k line app with composer 2.5 and grok 4.5. I love it", "context": "", "use_case": "", "switched_from": "", "story": "Cursor removed cost information from the usage page and CSV export"}, {"id": 783919617, "item_id": 48994976, "url": "https://news.ycombinator.com/item?id=48994976", "by": "gokuljs", "date": "2026-07-21", "version": null, "stance": "recommend", "quote": "I use Cursor and am happy with Composer :)", "excerpt": "I use Cursor and am happy with Composer :)", "context": "", "use_case": "AI coding model", "switched_from": "", "story": "Ask HN: Claude Code or Codex?"}, {"id": 783788162, "item_id": 48986760, "url": "https://news.ycombinator.com/item?id=48986760", "by": "striking", "date": "2026-07-20", "version": null, "stance": "recommend", "quote": "my Composer example speaks for itself", "excerpt": "I'm arguing we can't trust retail prices because the marginal pricing isn't meaningfully connected to it anyway. But if we have to look at what we think margins might look like, DeepSeek continues to host v4 Flash at the existing price despite competitors beating it in price (https://openrouter.ai/deepseek/deepseek-v4-flash), so there's at least one example of a Chinese lab charging a predetermined price despite comp", "context": "", "use_case": "", "switched_from": "", "story": "Who's Afraid of Chinese Models?"}, {"id": 783782241, "item_id": 48986390, "url": "https://news.ycombinator.com/item?id=48986390", "by": "striking", "date": "2026-07-20", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 is itself a fine-tune of K2.7, and it's apparently quite token efficient", "excerpt": "Sure, let's have a look... > I highly doubt that Chinese models are cheaper to serve on a marginal cost basis, they just seem cheaper because Anthropic and OpenAI are so supply constrained that they are charging far more than they would if there were sufficient supply to meet the demand for intelligence. [emphasis mine] I guess I'm missing the part of this article where they bring hard numbers in to back up the argum", "context": "", "use_case": "", "switched_from": "", "story": "Who's Afraid of Chinese Models?"}, {"id": 781973144, "item_id": 48873321, "url": "https://news.ycombinator.com/item?id=48873321", "by": "gabriel-uribe", "date": "2026-07-11", "version": null, "stance": "recommend", "quote": "Like GLM, (now with 4.5) Grok, Composer.", "excerpt": "Fwiw, don't buy into all the hype that you're falling behind. Yes, AI does cool things now, but I would say the impact is still unproven past indie hackers or early-stage startups. And a lot of the esoteric setups people have created with things like OpenClaw have become outdated as quickly as they were conceived. The popular thing is now to setup loops (eg I setup hourly integrations for Claude/Codex to 1) scrape my", "context": "", "use_case": "reviewing model output", "switched_from": "", "story": "AI 2040: Plan A"}, {"id": 781862192, "item_id": 48866387, "url": "https://news.ycombinator.com/item?id=48866387", "by": "arendtio", "date": "2026-07-10", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 was a lot better than I had expected", "excerpt": "I also like to take a look at https://cursor.com/cursorbench While in the past months Composer 2.5 was a lot better than I had expected a year ago, and the GPT 5.6 family does a good job in terms of cost for performance, I wonder why nobody is talking about Grok 4.5 high? Those numbers look very convincing to me.", "context": "", "use_case": "coding", "switched_from": "", "story": "GPT-5.6"}, {"id": 781503056, "item_id": 48843941, "url": "https://news.ycombinator.com/item?id=48843941", "by": "espadrine", "date": "2026-07-09", "version": "2.5", "stance": "neutral", "quote": "Composer 2.5 finetuned Kimi K2.5", "excerpt": "Composer 2.5 finetuned Kimi K2.5[0]. In the blog post, it is unclear whether Grok 4.5 is also a finetune on top of Kimi; they do imply it is also a finetune. > Training included trillions of tokens of Cursor data\u2026 We used reinforcement learning on difficult problems If xAI pivoted from a frontier base model company, to a finetuning company, it does mark a stark change to their relevance in the industry. [0]: https://", "context": "", "use_case": "", "switched_from": "", "story": "Grok 4.5"}, {"id": 781395712, "item_id": 48837232, "url": "https://news.ycombinator.com/item?id=48837232", "by": "qingcharles", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Composer 2.5 is worse than both; I use it all day for simple stuff.", "excerpt": "Composer 2.5 is worse than both; I use it all day for simple stuff. It's Kimi 2.6 in a new outfit.", "context": "", "use_case": "simple coding tasks", "switched_from": "", "story": "SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence"}, {"id": 781370464, "item_id": 48835654, "url": "https://news.ycombinator.com/item?id=48835654", "by": "bhouston", "date": "2026-07-08", "version": "2.5", "stance": "warn", "quote": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes.", "excerpt": "Every time I use Composer 2.5 I have to spend a bunch of time cleaning up its mistakes. It is unusable compared to GPT 5.4 or 5.5. My time is more valuable that I will use a model that doesn\u2019t f** up my code base.", "context": "", "use_case": "AI coding", "switched_from": "", "story": "Grok 4.5"}, {"id": 781349233, "item_id": 48834327, "url": "https://news.ycombinator.com/item?id=48834327", "by": "pants2", "date": "2026-07-08", "version": "2.5", "stance": "compare", "quote": "Kinda funny that their \"cost-vs-performance\" chart looks the same as the one for Composer 2.5", "excerpt": "Kinda funny that their \"cost-vs-performance\" chart looks the same as the one for Composer 2.5[1], except that it includes Composer 2.5 at a completely different spot. What are the chances that CursorBench ranks Cursor's model highest, and Cognition's bench ranks Cognition's model highest? Both are to be RL'd from Kimi as a base model, BTW. I'd posit that it's not deliberate deception, but for both companies their tra", "context": "", "use_case": "", "switched_from": "", "story": "SWE-1.7 Reach Near GPT 5.5 and Opus Intelligence"}, {"id": 781307378, "item_id": 48831711, "url": "https://news.ycombinator.com/item?id=48831711", "by": "jarodrh", "date": "2026-07-08", "version": "2.5", "stance": "neutral", "quote": "other models (GPT*, GEMINI*, Cursor - Composer 2.5 etc) as more like workhorse models", "excerpt": "Hmm, I honestly think a harness should be more about workflow, control, ease of use, memory optimization...other things I can't think of right now, the model/subscription being the least of them. Most of my experience has come from using Claude Code so I can only speak from that, but I will at some point pick up 'Pi' when I'm feeling adventurous. I feel completely free to use any number of subscriptions alongside Cla", "context": "", "use_case": "workhorse model for cost", "switched_from": "", "story": "Ask HN: What is your AI harness that lets you switch LLM models easily?"}, {"id": 780997952, "item_id": 48812372, "url": "https://news.ycombinator.com/item?id=48812372", "by": "richkuo", "date": "2026-07-06", "version": "2.5", "stance": "recommend", "quote": "Incredibly token efficient, fast, and gets the same quality as if you worked the whole thing with Fable/Opus/GPT.", "excerpt": "Composer 2.5 is really effective at some tasks, but doesn't do as well on higher complexity tasks from my own experience. With that being said, Cursor Bench is nonetheless a fantastic gauge of LLM quality and performance. The most interesting outcome of Cursor Bench is Fable 5 to GPT 5.5, its almost a perfect continuous line. As of the report, Fable to GPT might be the new standard of agentic programming/building I d", "context": "open source maintainer", "use_case": "building code after planning", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780430739, "item_id": 48776921, "url": "https://news.ycombinator.com/item?id=48776921", "by": "jarodrh", "date": "2026-07-03", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 (brilliant and fast)", "excerpt": "Orchestration works very well for me, but not in the way most people seem to be pushing for, with middlemen scoring and routing every request. I'm not sure I agree with that direction. For coding agents the routing problem is mostly solved at the config level, the harness already lets you pin models per role (done via subagents), and that covers most of what a per-request router promises. Three things have happened t", "context": "", "use_case": "agent pool coding", "switched_from": "", "story": "Ask HN: Is anyone experimenting with different ways of using LLMs for coding?"}, {"id": 780313664, "item_id": 48769604, "url": "https://news.ycombinator.com/item?id=48769604", "by": "Losenok", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow.", "excerpt": "Very skeptical about the composer accuracy. I have been using it for 6 months now and it is very fast, especially compared to anthropic models, but the result it produces, especially with more difficult tasks is very shallow. It feels like it just finds the cheapest way to deliver the task.", "context": "using it 6 months", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780269488, "item_id": 48766843, "url": "https://news.ycombinator.com/item?id=48766843", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "my distaste for Composer and my preference for Opus", "excerpt": "> it tends to leave big, dangerous holes hiding inside implementations unless babied it's fascinating that I used these same exact words to express my distaste for Composer and my preference for Opus. I suspect, the domains and problems we are trying to solve need to be shared. I wrote about it here: https://news.ycombinator.com/item?id=48766275 Would love to reach out to discuss more, if you're ok with it, or absolu", "context": "", "use_case": "AI coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780265537, "item_id": 48766596, "url": "https://news.ycombinator.com/item?id=48766596", "by": "subhobroto", "date": "2026-07-02", "version": null, "stance": "warn", "quote": "Composer hands me a quickly put together college project that was started the night before it was due.", "excerpt": "I'm not disputing what you're saying. The slowness of Opus in particular is pretty accurate but you should have been getting little popups in Cursor saying Opus is under load and to try switching to other models? I mentioned my frustration with Composer in another thread and why I rely on Opus, but Opus, atleast via Cursor is practically unusable for me M-F 9-5 EST. As a result, I have modified my working schedule ou", "context": "", "use_case": "AI coding via Cursor", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780262112, "item_id": 48766382, "url": "https://news.ycombinator.com/item?id=48766382", "by": "soyin", "date": "2026-07-02", "version": "2.5", "stance": "recommend", "quote": "I find it really hard to justify the other models for the performance/cost I'm getting with Composer 2.5.", "excerpt": "I'm also using it as my daily driver. I've been trying Opus 4.8 this week to see if I was missing something but haven't noticed a meaningful difference. I'm working on a fairly routine full stack web app that isn't doing anything incredible. Once I had the patterns I wanted in place, it's been very capable of following those with new work. I also don't ever give it long running tasks, it's always focused and small ch", "context": "working on a fairly routine full stack web app", "use_case": "full stack web app coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780142192, "item_id": 48758887, "url": "https://news.ycombinator.com/item?id=48758887", "by": "maxdo", "date": "2026-07-02", "version": "2.5", "stance": "recommend", "quote": "I use for everything like trigger and monitor a 10 steps release process using composer , a very capable model", "excerpt": "The opposite , I use for everything like trigger and monitor a 10 steps release process using composer , a very capable model", "context": "", "use_case": "release process automation", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780119777, "item_id": 48757486, "url": "https://news.ycombinator.com/item?id=48757486", "by": "mdasen", "date": "2026-07-01", "version": "2.5", "stance": "warn", "quote": "Artificial Analysis' testing shows Composer 2.5 to be pretty far behind", "excerpt": "I'm a bit skeptical. Cursor's benchmark finds that Cursor's model (Composer 2.5) is basically as good as Opus 4.8 max and GPT-5.5 xhigh, but at a fraction of the price. Artificial Analysis' testing shows Composer 2.5 to be pretty far behind: https://artificialanalysis.ai/agents/coding-agents. You look at the DeepSWE benchmark (which is probably the hardest to game at this point) and GPT-5.5 xhigh gets a 64, Opus 4.8 ", "context": "", "use_case": "", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 780117184, "item_id": 48757324, "url": "https://news.ycombinator.com/item?id=48757324", "by": "danfritz", "date": "2026-07-01", "version": "2.5", "stance": "recommend", "quote": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done.", "excerpt": "It's my daily driver, it's fast affordable and with a bit of guidance gets the job done. I only reach for Claud when i need to plan something big or want to have a sparring partner to fire of some ideas. I think what a lot of people don't realize is that you don't need a fronteer model for 80% of coding tasks. Composer 2.5 is often more than good enough, less token hungry and way faster", "context": "", "use_case": "daily coding", "switched_from": "", "story": "CursorBench 3.1"}, {"id": 779824403, "item_id": 48739025, "url": "https://news.ycombinator.com/item?id=48739025", "by": "ifwinterco", "date": "2026-06-30", "version": null, "stance": "recommend", "quote": "I use Composer (since we use Cursor) or GPT 5.3-codex as my workhorse models", "excerpt": "I prefer GPT 5.5 to Opus but both are absurdly expensive token hogs, I can't afford to use either as my main model at $work with the monthly spend cap we have. I use Composer (since we use Cursor) or GPT 5.3-codex as my workhorse models and only break out the big guns when I have a genuinely difficult problem to solve. IMO somewhat weirdly 5.3-codex might be the best overall coding model OpenAI have ever released. It", "context": "works under a monthly spend cap at work", "use_case": "main coding workhorse model", "switched_from": "", "story": "Claude Sonnet 5"}, {"id": 779792880, "item_id": 48737055, "url": "https://news.ycombinator.com/item?id=48737055", "by": "AdminAdmim", "date": "2026-06-30", "version": "2.5", "stance": "recommend", "quote": "when i use Cursor i use 90% Composer 2.5 fast", "excerpt": "Same for me, downgraded Cursor Subscription because when i use Cursor i use 90% Composer 2.5 fast", "context": "", "use_case": "Cursor coding model", "switched_from": "", "story": "Claude Sonnet 5"}, {"id": 777288482, "item_id": 48580530, "url": "https://news.ycombinator.com/item?id=48580530", "by": "tomerbd", "date": "2026-06-17", "version": "2.5", "stance": "neutral", "quote": "composer 2.5 - mediocre with hand holding and super fast", "excerpt": "I code daily with AI - real programming tasks, real work I use below 3: codex 5.5 medium - best results less hand holding glm 5.2 max - mediocre with hand holding and super slow composer 2.5 - mediocre with hand holding and super fast I use all, since i run mulitple coding in parallel. disclosure - I use rexide which we created for all these agents to run in parallel with good visibility and feedback.", "context": "codes daily with AI", "use_case": "AI coding", "switched_from": "", "story": "GLM-5.2 is the new leading open weights model on Artificial Analysis"}, {"id": 777084561, "item_id": 48567785, "url": "https://news.ycombinator.com/item?id=48567785", "by": "jemmyw", "date": "2026-06-17", "version": null, "stance": "recommend", "quote": "Composer is actually really good, and fast.", "excerpt": "I didn't like cursor when it first came about but now I use it for my personal projects. The plan is good value for accessing different sota models occasionally for planning. Composer is actually really good, and fast. I'm not sure if it's because my personal projects are small enough I know them inside out and my work project is huge, but I prefer terminal code agents for day work over ide integration.", "context": "", "use_case": "", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 777047041, "item_id": 48565440, "url": "https://news.ycombinator.com/item?id=48565440", "by": "bottlepalm", "date": "2026-06-16", "version": null, "stance": "recommend", "quote": "", "excerpt": "Can't remember the last time I used AI tab complete; back in the day indeed.", "context": "", "use_case": "AI coding model", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 777034048, "item_id": 48564628, "url": "https://news.ycombinator.com/item?id=48564628", "by": "ghshephard", "date": "2026-06-16", "version": null, "stance": "recommend", "quote": "that I use a lot of the time when I have fairly straightforward tasks", "excerpt": "You nailed it - in fact, most of Anthropic's early revenue came from Cursor - much of claude code programming components is essentially a feature copy of Cursor, so it makes sense they are similar. Cursor does have it's own model - it's a heavily reworked version of KimiK2, called \"composer\" - that I use a lot of the time when I have fairly straightforward tasks that don't require a lot of exploration or independent ", "context": "", "use_case": "straightforward coding tasks", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 777000067, "item_id": 48562504, "url": "https://news.ycombinator.com/item?id=48562504", "by": "armcat", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "mirror what Cursor has done with Composer 2.5", "excerpt": "I keep seeing these \"sovereign\" LMs time and time again. In Sweden we had GPT-SW3 (https://www.ai.se/en/project/gpt-sw3) and same story there. Instead of burning money on \"sovereign\" claims, national research labs should instead focus on building on top of solid baselines (like Qwen/Kimi) and finetuning frontier models with real agentic utility that can be applied across actual use cases and can be widely used by its", "context": "", "use_case": "", "switched_from": "", "story": "GPT\u2011NL: a sovereign language model for the Netherlands"}, {"id": 776963841, "item_id": 48560240, "url": "https://news.ycombinator.com/item?id=48560240", "by": "darklinear", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 is really fast in their harness at code search/explanation/Q&A tasks", "excerpt": "It's still my primary at work (data engineering/platform engineering), running a mix of GPT-5.5 and Composer 2.5. We also have Claude Code subscriptions. I find myself preferring Cursor for most tasks. 70% of the time I use AI agents on a pretty tight leash. I often reject edits and ask it to change things. The IDE integration is really efficient for this workflow compared to Claude (yes, I've tried the Claude extens", "context": "data engineering/platform engineering", "use_case": "code search and debugging", "switched_from": "", "story": "SpaceX Is Buying Cursor"}, {"id": 776949553, "item_id": 48559347, "url": "https://news.ycombinator.com/item?id=48559347", "by": "TrisMcC", "date": "2026-06-16", "version": "2.5", "stance": "neutral", "quote": "mostly using Composer 2.5, but I freely change the model when the need arises", "excerpt": "I use cursor (through a work subscription), only the cli (https://cursor.com/cli), and mostly using Composer 2.5, but I freely change the model when the need arises. Most comments here seem to think there is no command line client? I have never used the editor. For my personal projects, I use a heavily modified pi. I also have access to a claude code account through work (bedrock), but I don't use it much. It always ", "context": "", "use_case": "coding model", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 776893616, "item_id": 48555851, "url": "https://news.ycombinator.com/item?id=48555851", "by": "infecto", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "I use it quite a bit and it works well and is extremely fast for the tasks it was built for", "excerpt": "Hmmm. Not in my experience. I don\u2019t think it can be compared to Haiku, maybe sonnet levels? It\u2019s obviously not Opus and never was intended for that use case. I use it quite a bit and it works well and is extremely fast for the tasks it was built for", "context": "", "use_case": "fast coding tasks", "switched_from": "", "story": "SpaceX Is Buying Cursor"}, {"id": 776868512, "item_id": 48554282, "url": "https://news.ycombinator.com/item?id=48554282", "by": "senordevnyc", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "I agree, Composer 2.5 is really good.", "excerpt": "I agree, Composer 2.5 is really good. I use it for all kinds of small tasks, and really for any kind of first pass at debugging, answering questions about the codebase, pulling data for reports, etc. It\u2019s fast, pretty accurate, and basically free.", "context": "", "use_case": "small tasks, debugging, codebase questions", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 776868065, "item_id": 48554254, "url": "https://news.ycombinator.com/item?id=48554254", "by": "senordevnyc", "date": "2026-06-16", "version": "2.5", "stance": "recommend", "quote": "Composer 2.5 is really solid", "excerpt": "Yes, it\u2019s my daily driver for building the saas I run full-time. I\u2019m not happy about this news. I like the ability to switch between any models, Composer 2.5 is really solid, I like having my agents coworking in the IDE with me, the plan mode is great, Cloud Agents are great, especially with slack, linear, web, etc integrations. I routinely tag an error report in slack and Cursor fires up a Composer 2.5 cloud agent t", "context": "runs a SaaS full-time", "use_case": "cloud agent issue triage", "switched_from": "", "story": "SpaceX to buy Cursor AI coding agent operator Anysphere for $60B"}, {"id": 776119586, "item_id": 48507474, "url": "https://news.ycombinator.com/item?id=48507474", "by": "ignoramous", "date": "2026-06-12", "version": "2", "stance": "compare", "quote": "Composer 2 performed differently on evals than Moonshot.ai's coding models", "excerpt": "> Cursor had a specific licensing agreement... Cursor had an \"agreement\" with Fireworks.ai, which apparently allowed them to RL Composer 2 atop Kimi Base 2.5 without attribution: https://x.com/Kimi_Moonshot/status/2035074972943831491 / https://archive.vn/CcdkI Composer 2 performed differently on evals than Moonshot.ai's coding models: Cursor claims theirs is better than Claude Opus 4.6: https://x.com/fynnso/status/20", "context": "", "use_case": "", "switched_from": "", "story": "Kimi K2.7-Code: open-source coding model with better token efficiency"}, {"id": 772861824, "item_id": 48303864, "url": "https://news.ycombinator.com/item?id=48303864", "by": "RevEng", "date": "2026-05-27", "version": "2", "stance": "recommend", "quote": "I use composer-2 daily for complex programming tasks.", "excerpt": "I use composer-2 daily for complex programming tasks. It's a fine tuned Kimi 2.5 - nothing groundbreaking. I've even had reasonable success using Qwen 3.5 on my desktop GPU. Opus might be better, but it's certainly not necessary to get good results.", "context": "", "use_case": "Complex programming tasks", "switched_from": "", "story": "I think Anthropic and OpenAI have found product-market fit"}, {"id": 772437633, "item_id": 48277352, "url": "https://news.ycombinator.com/item?id=48277352", "by": "59nadir", "date": "2026-05-26", "version": "2", "stance": "warn", "quote": "They really can't produce much more than a function or a small module without shitting the bed", "excerpt": "My personal experience with trying to front-load tons of planning and speccing out with LLMs is that at best it's a small improvement on code quality but with considerably more time spent. As a result I've abandoned the idea of having LLMs generate code except for very small, localized and tightly scoped things. They really can't produce much more than a function or a small module without shitting the bed (last time ", "context": "", "use_case": "", "switched_from": "", "story": "Using AI to write better code more slowly"}, {"id": 771150369, "item_id": 48196898, "url": "https://news.ycombinator.com/item?id=48196898", "by": "wunderlotus", "date": "2026-05-19", "version": "2", "stance": "warn", "quote": "I've experienced Composer 2's performance and it leaves much to be desired as a daily driver for a knowledge worker.", "excerpt": "I love Cursor as a tool, but I'm skeptical bc: 1/ CursorBench is so opaque [1] that it makes it hard to trust. Not to mention the v3.1 eval is a newer iteration and there's no insight into the tasks or if the model was just tuned to max it out. Composer 2 previously scored between 60-65% on the previous benchmark eval [2] but scores between 50-55% on CB v3.1[3]. 2/ I've experienced Composer 2's performance and it lea", "context": "knowledge worker", "use_case": "daily driver coding model", "switched_from": "", "story": "Cursor Introduces Composer 2.5"}, {"id": 769211072, "item_id": 48075692, "url": "https://news.ycombinator.com/item?id=48075692", "by": "RevEng", "date": "2026-05-09", "version": "2", "stance": "neutral", "quote": "I use composer-2 and sonnet 4.6 on a regular basis", "excerpt": "Except that coding agents will do this at times. That's half the problem. A human will forget details and exaggerate others, but LLMs fail in spectacular ways that humans rarely would, like trying to copy a document from memory rather than one word at a time, side by side, or rewriting the whole thing just to make some simple changes. Coding agents will delete tests or return True to get them to pass - something you ", "context": "", "use_case": "", "switched_from": "", "story": "LLMs Corrupt Your Documents When You Delegate"}, {"id": 765814144, "item_id": 47863384, "url": "https://news.ycombinator.com/item?id=47863384", "by": "dmix", "date": "2026-04-22", "version": "2", "stance": "recommend", "quote": "I used to hate on Composer 2 but I'm coming around to it.", "excerpt": "I used to hate on Composer 2 but I'm coming around to it. Opus for the big stuff and multi-file operations, Composer for all the small day-to-day IDE tasks works pretty good for me.", "context": "", "use_case": "", "switched_from": "", "story": "SpaceX and Cursor partnership. Right to acquire Cursor later this year"}, {"id": 765777744, "item_id": 47861109, "url": "https://news.ycombinator.com/item?id=47861109", "by": "mzl", "date": "2026-04-22", "version": "2", "stance": "recommend", "quote": "Composer-2 seems very useful in Cursor", "excerpt": "Composer-2 is based on Kimi K2.5, but with extensive RL. Cursor estimated 3x more compute on their RL than the original K2.5 training run (some details in https://cursor.com/blog/composer-2-technical-report). Composer-2 seems very useful in Cursor, while K2.6 according to AA seems to be a really useful general model: https://artificialanalysis.ai/articles/kimi-k2-6-the-new-leading-open-weights-model", "context": "", "use_case": "", "switched_from": "", "story": "SpaceX and Cursor partnership. Right to acquire Cursor later this year"}, {"id": 765744272, "item_id": 47859017, "url": "https://news.ycombinator.com/item?id=47859017", "by": "JacobWolf", "date": "2026-04-21", "version": "2", "stance": "recommend", "quote": "Composer 2 is legitimately so impressive", "excerpt": "Yeah, Composer 2 is legitimately so impressive. It is my daily driver right now both on professional and personal projects. I only find myself reaching for 5.3 Codex/GPT 5.4 when exploring a lot of technical documentation or code and for Sonnet/Opus when working on UI. Everything else is Composer.", "context": "", "use_case": "", "switched_from": "", "story": "SpaceX and Cursor partnership. Right to acquire Cursor later this year"}], "latest_version": {"label": "Cursor Composer 2.5", "people": 35, "rec": 23, "warn": 3, "other": 9}}