Skip to content

Benchmark

SkillsBench 1.1 · With Skills

Scores are percentages. Higher is better.

Models and systems measured
26
Scores recorded
30
Sources
1
IQ weight
2.24

Score distribution

52.6%
Mean
53.1%
Median
12.2%
Std dev
36.2%
10th percentile
67.3%
90th percentile
20.1%
Min
20.1%73.5%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.24 for chat models, measured over 24 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1GPT-5.6 SolQwen他测:v1.1 · 87 tasks · 每题3次均值;Opus/Fable=Claude Code,GPT=Codex,Qwen=OpenCode· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.73.5%
2Claude Fable 5Qwen他测:v1.1 · 87 tasks · 每题3次均值;Opus/Fable=Claude Code,GPT=Codex,Qwen=OpenCode· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.9%
3Qwen3.8 MaxQwen他测:v1.1 · 87 tasks · 每题3次均值;Opus/Fable=Claude Code,GPT=Codex,Qwen=OpenCode· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70.2%
4GPT 5.5OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs67.3%
5GPT 5.5Codex · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs66.5%
6Claude Opus 4.8Qwen他测:v1.1 · 87 tasks · 每题3次均值;Opus/Fable=Claude Code,GPT=Codex,Qwen=OpenCode· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.65.1%
7Claude Opus 4.7Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs61.2%
8Qwen3.7 MaxQwen他测:v1.1 · 87 tasks · 每题3次均值;Opus/Fable=Claude Code,GPT=Codex,Qwen=OpenCode· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.61.2%
9Gemini 3.1 Pro PreviewGemini CLI · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs60.8%
10GLM 5.1OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs58.4%
11Hy3 (Deep Infra)Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs55.9%
12Gemini 3 FlashGemini CLI · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs54.6%
13Claude Opus 4.8OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs54.1%
14Kimi K2.6OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs54%
15Claude Opus 4.7OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs53.1%
16MiniMax M3OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs53%
17Gemini 3.1 Pro PreviewOpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs52.8%
18GPT 5.2Codex · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs51.7%
19Claude Opus 4.6Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs50.2%
20Claude Opus 4.5Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs49%
21Gemini 3.5 FlashOpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs48.2%
22Claude Sonnet 4.6OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs47.2%
23DeepSeek V4 FlashOpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs44.7%
24GPT-5.4 miniOpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs41.4%
25Claude Sonnet 4.5Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs36.2%
26MiniMax M2.7OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs34.9%
27Claude Haiku 4.5Claude Code · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs30.1%
28Gemini 3.1 Flash LiteOpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs20.1%

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
DeepSeek V4 Pro (version unspecified)OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs50.1%
Grok 4.3OpenHands · 87 tasks · 3 trials/task · max reasoning effort · with Skills· Reported by SkillsBench / BenchFlow· via BenchBoard docs41.7%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.