返回
RCreddit.com
18
·23小时前·RSS
暂不在当前实时榜单

I collected every single LLM coding benchmark, and computed their Intelligence Density

查看原文
开源代码

热度趋势

新上榜
最近 24 小时与此前 24 小时对比 · 7 天曲线

百分比基于当前可用热度信号,而非评论数或独立用户人数。

AI 摘要

An aggregate index called the Agentic Coding Index was computed across several agentic coding benchmarks, including SWE-bench Pro, DeepSWE v1.1, Terminal-Bench (v4, v3, v2.1), Code Arena Elo, and LiveCodeBench v6. DeepSWE v1.1 and Code Arena Elo each contribute 20% to this index, with Terminal-Bench v4.0 and SWE-bench Pro contributing 15% each. Terminal-Bench v3.0, v2.1, and LiveCodeBench v6 contribute 13%, 12%, and 5% respectively. All benchmark scores are from verified public and official sources.