CursorBench更新私有評測
CostTokensStepsModelScoreCostCost / taskTokensTokens / taskStepsSteps / task1Fable 5.1 Max51.8%$17.28117,2361282Fable 5.1 Extra High51.6%$13.0187,2941013Fable 5.1 High49.2%$9.0858,438774Fable 5.1 Medium46.8%$7.0545,411635Opus 5 Max46.6%$11.9585,3841066Opus 5 Extra High46.1%$11.4380,0941037Fable 5.
1 Low45.1%$5.4434,795518Opus 5 High44.7%$9.0061,405869Opus 5 Medium43.3%$6.9445,2727210GPT-5.6 Sol Max41.7%$8.2342,9449911Muse Spark 1.3 Max41.6%$2.6452,0059812Grok 4.6 Extra High41.4%$6.1049,8145613GPT-5.6 Terra Max41.3%$5.1460,81410714Opus 5 Low40.7%$4.8731,9955715Grok 4.6 High40.4%$5.
2041,3874816Gemini 3.8 Flash High39.6%$4.70162,56532417GPT-5.6 Sol Extra High37.7%$4.4024,7295518Muse Spark 1.3 Extra High37.5%$2.1040,8918319Gemini 3.8 Flash Medium37.3%$4.06128,36429020Grok 4.6 Medium36.1%$3.4824,8934021GPT-5.6 Luna Max35.9%$1.0387,28420822GPT-5.6 Sol High35.7%$2.
8516,1744123Sonnet 5 Max34.1%$7.17149,25714024GPT-5.6 Terra Extra High33.6%$1.8123,4364325Grok 4.6 Low33.4%$2.2516,3073226Muse Spark 1.3 High33.4%$1.6630,6546927GPT-5.6 Luna Extra High33.0%$0.4440,5989828Muse Spark 1.3 Medium32.6%$1.4927,2556429Sonnet 5 Extra High32.0%$4.5583,37310230GPT-5.
6 Sol Medium31.1%$1.7710,1113231Sonnet 5 High30.8%$3.4861,1468532GPT-5.6 Terra High30.7%$1.1113,1623333GPT-5.6 Luna High29.4%$0.2523,3686434Muse Spark 1.3 Low29.3%$0.9317,4834735Sonnet 5 Medium28.0%$2.3139,1146536Composer 2.527.7%$0.6817,3474137GPT-5.6 Terra Medium27.6%$0.647,3072538GPT-5.
6 Terra Low25.2%$0.525,9142339GPT-5.6 Sol Low24.6%$0.874,8852140Muse Spark 1.3 Minimal24.3%$0.5610,6203441Sonnet 5 Low24.1%$1.3923,7724642GPT-5.6 Luna Medium22.2%$0.087,6423243GPT-5.6 Luna Low16.0%$0.033,28818ChangelogSep 10, 2026TasksCursorBench 4.
0Introduced new long-horizon problems focused on edit, refactor, investigation, intent understanding, managing jobs, and design adherence.Aug 11, 2026ReportingUpdated Sonnet 5 results to account for adjusted pricing.Jul 30, 2026ReportingUpdated GPT-5.
6 Terra and Luna results to account for adjusted pricing.Jul 9, 2026ReportingUpdated GPT-5.6 Sol, Terra, and Luna results to account for cache write costs.Jul 8, 2026TasksCursorBench 3.2Introduced instruction following and advanced tool use problems.May 19, 2026TasksCursorBench 3.
1Introduced problems focused on codebase understanding, bugfinding, planning, and code review.Improved grading criteria for some edit tasks.Mar 11, 2026TasksCursorBench 3.0Initial set of tasks focused on edit, refactor, and bugfix problems.
Avg cost / task is computed by applying each model's published per-million-token pricing (input, cache read, cache write, and output) to the tokens it used on each task.Results are subject to variance; small differences in scores may not be statistically meaningful.
Related
相關文章
無問芯穹與華環電子簽署戰略合作，共同探索國產異構算力AI基礎設施新方向
無問芯穹與華環電子簽署戰略合作協議，雙方將結合各自在AI軟體平台、網路通信與硬體研發的優勢，共同探索國產異構算力基礎設施的協同方案。此次合作聚焦於智算中心解決方案及「Token工廠」新模式，目標是推動計算、網路與AI原生基礎設施深度融合，為AI規模化應用提供高效穩定的支撐。
Jina AI Releases jina-ocr-v1: A 3.4B MoE Document Parser With Built-In Speculative Decoding for Low-Budget GPUs
Jina AI, part of Elastic, has released jina-ocr-v1, an end-to-end visual document parser. It takes PDFs, scans, tables, charts or invoices and returns clean Markdown in 1 pass. The model has 3.
優步全球範圍裁員 10%，被裁員工稱 AI 已大舉滲透日常工作
作者：清源 責編：清源 評論： 9 月 18 日消息，據《商業內幕》今天（18 日）晚間報道，在優步（Uber），AI 已經滲透到員工工作的許多環節，從回答 Slack 裡的內部問題，到替乘客行程中聯繫客服時收到的消息撰寫回復。6 名近期遭裁員的員工透露，過去幾個月，AI 在工作中的使用範圍明顯擴大，其中一些人甚至會通過提示詞讓 AI 完成相當一部分任務。
智譜 ZCode 被質疑“偷傳代碼”：官方回應稱問題已修復，將開源代碼庫、引入第三方審查
作者：清源 責編：清源 評論： 感謝網友 咩咩洋 的線索投遞！9 月 18 日消息，針對社區中有關代碼庫數據上傳的討論，智譜旗下編程產品 ZCode 今天（18 日）通過智譜官方群組向受影響用戶致歉，併發布回應稱已第一時間完成自查，相關問題目前已經修復。
月之暗面遞表之後，Kimi 的成色要被驗算三遍
舒澤品牌手記2026.09.18 18:16 · 來自浙江全文4982字00:00 / 14:05Anthropic 的 30 萬次指控，會成為招股書的第幾頁？文 | 舒澤品牌手記9月17日，月之暗面發佈了一套金融行業解決方案。按官方披露，中信建投、中金公司、易方達等數十家金融機構已經在用 Kimi 處理投研建模、風險排查和盡調材料——研究人員把管理層報表、審計報告和盡調文件交給 Kimi，拿回一份可以繼續調整假設的 Excel 模型。同一天，深圳商報記者就港股上市進展、股東架構調整等事項向月之暗面發去採訪函。
Calibre上手 AI 互動寫作：電子書管理器搖身變成"文字冒險遊戲引擎"
這個遊戲默認藏而不發，不會跟著 Calibre 啟動就冒出來。用戶得主動在"首選項 — 工具欄和菜單"裡把它請到主工具欄，才算真正激活。它的玩法很清晰:由 AI 在後臺搭起並掌管一個虛構世界，用戶通過不斷輸入文字來推著故事往前走，等於把"讀電子書"這件事，翻轉成了"和 AI 一起寫故事"。