[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-cursor-composer-2-5-79pct-swe-bench-price":3,"topics-all":36,"news-related-60dd497d-fb0b-428d-9915-475ff760d2f2":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"60dd497d-fb0b-428d-9915-475ff760d2f2","Cursor自研编程模型Composer 2.5：闯入第一梯队的价格屠夫","5月18日，Cursor发布了第三代自研编程模型Composer 2.5。核心数字相当震撼：SWE-Bench Multilingual 79.8%，与Claude Opus 4.7和GPT-5.5基本持平；在SWE-Bench-Pro-Hard-AA这个对顶级模型最具挑战性的子集上，Composer 2.5比上代提升了35个百分点。更值得关注的是价格：标准 tier 每百万token输入0.5美元、输出2.5美元，大约是顶级闭源模型的十分之一。这意味着什么？编程AI的经济账正在被改写。此前，SWE-Bench分数能超过75%的模型几乎只有Anthropic和OpenAI的旗舰产品，高昂的推理成本让很多团队在好用和用得起之间被迫二选一。Composer 2.5用十分之一的成本做到了同样的准确率，这对长时间运行的AI编程智能体来说意义重大。技术层面，Composer 2.5建立在Moonshot的Kimi K2.5基础之上。Cursor透露，85%的计算预算用于在基座之上做额外后训练和强化学习，合成任务数据量是上代的25倍。这种站在开源肩上加大规模专项调优的路径，正在成为中小型实验室对抗头部厂商的主流策略。Composer 2.5的发布代表编程AI智能体市场正式进入性价比竞争阶段，不再只是谁的分数最高的竞争。接下来要看的是：当入门级编程智能体的成本降到原来的十分之一，传统IDE插件市场和Copilot们的定价策略会受到多大压力？这场价格战的受益者最终是开发者。","https:\u002F\u002Fcursor.com\u002Fblog\u002Fcomposer-2-5","02de7800-ece3-4d55-9ce8-f036f72dd9d4",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"e676a5cf-1f24-472f-a765-86fa21a1bc3c","ai-model",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":18,"name":19,"slug":19,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":21,"name":22,"slug":22,"description":13,"color":13},"01598627-1ea6-4b27-a5d8-874971571a71","llm",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"cc8f00c8-9407-4d78-80af-f525554c3000","en","Cursor's Composer 2.5: first-tier coding model at a cut price","On May 18, Cursor released its third-generation in-house coding model, Composer 2.5. The core numbers are striking: 79.8% on SWE-Bench Multilingual, essentially on par with Claude Opus 4.7 and GPT-5.5; on the SWE-Bench-Pro-Hard-AA subset — the most challenging slice for top-tier models — Composer 2.5 jumped 35 percentage points over its predecessor. Even more noteworthy is the price: $0.50 per million input tokens and $2.50 per million output tokens on the standard tier — roughly one-tenth of top closed-source models. What does that mean? The economics of coding AI are being rewritten. Previously, only Anthropic and OpenAI's flagships had models above 75% on SWE-Bench, and the steep inference cost forced many teams to choose between \"good\" and \"affordable.\" Composer 2.5 hits the same accuracy at one-tenth the cost — a big deal for long-running AI coding agents.\n\nOn the technical side, Composer 2.5 is built on Moonshot's Kimi K2.5. Cursor revealed that 85% of its compute budget went into post-training and reinforcement learning on top of the base, with 25× more synthetic task data than the previous generation. This \"stand on the shoulders of open source and apply large-scale targeted tuning\" approach is becoming the mainstream playbook for small-to-mid labs competing against the giants.\n\nThe release of Composer 2.5 marks the formal entry of the AI coding-agent market into a price-performance competition — no longer just \"whose score is highest.\" The question now is: with the cost of entry-level coding agents dropping to one-tenth, how much pressure will that put on the pricing strategies of traditional IDE plugins and the Copilots? The ultimate winners of this price war are the developers.","cursor-composer-2-5-79pct-swe-bench-price","2026-05-25T07:10:00Z","2026-05-25T07:06:43.412650Z","2026-08-19T02:08:40.142862Z",true,"agent",199,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"4bbc55d2-cabc-477f-a3ad-4e2c119aff2a","TokTier 抓住 Agent 推理的隐藏瓶颈：缓存命中 94.1%，分词仍吃掉 64% 首 token 时间","toktier-stateful-tokenization-agent-serving","2026-07-31T17:56:30+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"6b52b4a9-d567-46b8-99c1-e9c65ba59b16","SWE-Pruner Pro:ByteDance 让 Agent 自己当剪枝器,省 39% token 还涨分","swe-pruner-pro-bytedance","2026-07-25T12:00:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"3191469f-3250-4b31-bc42-8b43680a61a4","MeMo：把 LLM 的『记忆』和『推理』彻底拆开，更新知识再也不用重训","memo-memory-as-model-90pct-compute-saved","2026-06-19T06:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"436fb2b9-4c48-4631-977c-c9539650f975","Kimi K2.7-Code 开源:Moonshot 把\"过度思考\"砍掉三成,长程编程更经济","kimi-k2-7-code-moonshot-30pct-token-cut","2026-06-13T02:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"4326dbe6-f7c1-4ce8-8ef1-8cd7aa1cbb97","ReLoRA：基础模型频繁更新下的LoRA适配器复用之道","relora-base-model-lora-reuse-89pct-time","2026-06-03T08:10:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"b4a4434f-6b1e-4ff3-ae69-2d9e82ab3e29","小米MiMo-V2.5首度揭秘：五大推理优化技术如何实现「降价不亏本」","xiaomi-mimo-v2-5-five-inference-optimization","2026-05-31T04:00:00+00:00"]