请问公开的基准测试是否有指定语言的细分数据？

[Grey4sh]

Problem Statement

比如我想知道 Go 语言那些测试题在 OCR 上的表现

Proposed Solution

/

Alternatives Considered

/

Affected Area

Documentation

Additional Context

No response

[stay-foolish-forever]

Hi @Grey4sh，感谢关注！

以下是目前 Go 语言上的基准测试指标：

#	模型	Agent	F1	精确率	召回率
1	Claude-4.6-Opus	Open Code Review	27.21%	33.33% (40/120)	22.99% (40/174)
2	GPT-5.5	Open Code Review	26.22%	37.63% (35/93)	20.11% (35/174)
3	GLM-5.2	Open Code Review	25.00%	39.02% (32/82)	18.39% (32/174)
4	GLM-5.1	Open Code Review	22.71%	31.31% (31/99)	17.82% (31/174)
5	Claude-4.8-Opus	Open Code Review	21.01%	39.06% (25/64)	14.37% (25/174)
6	Qwen3.7-Max	Open Code Review	19.60%	19.65% (34/173)	19.54% (34/174)
7	Claude-4.8-Opus	Claude Code	19.35%	19.76% (33/167)	18.97% (33/174)
8	Deepseek-V4-Pro	Open Code Review	17.36%	25.27% (23/91)	13.22% (23/174)
9	GPT-5.5	Codex	12.26%	34.21% (13/38)	7.47% (13/174)
10	Deepseek-V4-Pro	Claude Code	11.81%	8.98% (30/334)	17.24% (30/174)
11	Qwen3.7-Max	Claude Code	10.67%	7.00% (39/557)	22.41% (39/174)
12	Claude-4.6-Opus	Claude Code	10.22%	6.19% (51/824)	29.31% (51/174)
13	GLM-5.1	Claude Code	10.00%	6.48% (38/586)	21.84% (38/174)

我们后续会在公开的 Leaderboard 中维护 Bench 上各语言维度的细分指标，届时可以直接在 Leaderboard 页面查看不同语言的表现数据。

[Grey4sh]

非常感谢！

[Grey4sh]

@stay-foolish-forever 请问榜单里的 GPT 5.5 模型在评测时thinking effort 开的是 xhigh吗？