@estebankiwiCursor’s proprietary Composer 2.5 model has been released, and its benchmarks are nearly indistinguishable from Anthropic’s Claude Opus 4.7. On Terminal-Bench 2.0, Composer 2.5 scored 69.3% versus Opus 4.7’s 69.4%. In SWE-Bench Multilingual, the gap was similarly narrow: 79.8% for Composer against 80.5% for Opus. On CursorBench v3.1, Composer 2.5 also performed competitively. The company has essentially built a model that goes head-to-head with Anthropic’s offering.
查看原帖@estebankiwi 的照片· May 18, 2026· Anthropic
The focus is on a comparison table featuring different software versions. It displays performance percentages for Composer 2.5, Opus 4.7, GPT-5.5, and Composer 2 across three benchmarks: Terminal-Bench 2.0, SWE-Bench Multilingual, and CursorBench v3.1. The style is data-driven and analytical, with a simple white background and black text. Notably, Composer 2.5 is highlighted with a red border, drawing attention to its performance metrics.
Jacob Coxon AI warning
2Jacob Coxon Anthropic resignation
2Former OpenAI Anthropic researcher resigns over AI extinction fears
Fable 5.1 job offer
Anthropic AI extinction odds
Anthropic engineer explains Claude Code workflow
2US government Anthropic trust
2Anthropic Fable 5.1 Mythos 5.1
2Anthropic First Amendment ruling
2Anthropic Claude lab operator
Anthropic Claude Academy launch
2Anthropic IPO Claude AI
Claude vs ChatGPT free speech debate
Isabella He Anthropic Code w/ Claude@estebankiwiCursor’s proprietary Composer 2.5 model has been released, and its benchmarks are nearly indistinguishable from Anthropic’s Claude Opus 4.7. On Terminal-Bench 2.0, Composer 2.5 scored 69.3% versus Opus 4.7’s 69.4%. In SWE-Bench Multilingual, the gap was similarly narrow: 79.8% for Composer against 80.5% for Opus. On CursorBench v3.1, Composer 2.5 also performed competitively. The company has essentially built a model that goes head-to-head with Anthropic’s offering.
查看原帖@estebankiwi 的照片· May 18, 2026· Anthropic
The focus is on a comparison table featuring different software versions. It displays performance percentages for Composer 2.5, Opus 4.7, GPT-5.5, and Composer 2 across three benchmarks: Terminal-Bench 2.0, SWE-Bench Multilingual, and CursorBench v3.1. The style is data-driven and analytical, with a simple white background and black text. Notably, Composer 2.5 is highlighted with a red border, drawing attention to its performance metrics.
还没有评论。来抢沙发吧！
Jacob Coxon AI warning2Jacob Coxon Anthropic resignation2Former OpenAI Anthropic researcher resigns over AI extinction fearsFable 5.1 job offerAnthropic AI extinction oddsAnthropic engineer explains Claude Code workflow2US government Anthropic trust2Anthropic Fable 5.1 Mythos 5.12Anthropic First Amendment ruling2Anthropic Claude lab operatorAnthropic Claude Academy launch2Anthropic IPO Claude AIClaude vs ChatGPT free speech debateIsabella He Anthropic Code w/ Claude
还没有评论。来抢沙发吧！