Skip to content

Benchmark

CL-bench Life · Overall

Scores are percentages. Higher is better.

Models and systems measured
23
Scores recorded
47
Sources
2
IQ weight
1.44

Score distribution

13.2%
Mean
13.2%
Median
4.4%
Std dev
7.4%
10th percentile
19.3%
90th percentile
6.3%
Min
6.3%22.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.44 for chat models, measured over 15 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 5.5benchmark-owner leaderboard run · site model-mode label: GPT 5.5 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.5 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs22.2%
2GPT 5.5high· Epoch AI Benchmarking Hub data22.2%
3GPT 5.4benchmark-owner leaderboard run · site model-mode label: GPT 5.4 (xHigh) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs21.7%
4GPT 5.4xhigh· Epoch AI Benchmarking Hub data21.7%
5GPT 5.4benchmark-owner leaderboard run · site model-mode label: GPT 5.4 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.5 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs19.3%
6GPT 5.4high· Epoch AI Benchmarking Hub data19.3%
7GPT 5.1high· Epoch AI Benchmarking Hub data17.3%
8Claude Opus 4.6benchmark-owner leaderboard run · site model-mode label: Claude Opus 4.6 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±1.3 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs17%
9Claude Opus 4.6high· Epoch AI Benchmarking Hub data17%
10Hy3 (Deep Infra)Hy3 official Benchmark Appendix · highest reasoning tier unless noted· Reported by Tencent Hunyuan· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.17%
11Gemini 3.1 Pro Preview· Epoch AI Benchmarking Hub data16.9%
12Gemini 3.1 Pro Previewbenchmark-owner leaderboard run · site model-mode label: Gemini 3.1 Pro (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±1.2 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs16.9%
13Gemini 3.1 Pro Previewbenchmark-owner leaderboard run · site model-mode label: Gemini 3.1 Pro · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs16%
14GPT 5.4· Epoch AI Benchmarking Hub data13.8%
15GPT 5.4benchmark-owner leaderboard run · site model-mode label: GPT 5.4 · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs13.8%
16Claude Opus 4.6· Epoch AI Benchmarking Hub data13.6%
17Claude Opus 4.6benchmark-owner leaderboard run · site model-mode label: Claude Opus 4.6 · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs13.6%
18DeepSeek V4 Prohigh· Epoch AI Benchmarking Hub data13.5%
19Kimi K2.5· Epoch AI Benchmarking Hub data13.2%
20Kimi K2.5benchmark-owner leaderboard run · site model-mode label: Kimi K2.5 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.8 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs13.2%
21Qwen3.5 Plus· Epoch AI Benchmarking Hub data12.4%
22Grok 4.20· Epoch AI Benchmarking Hub data11.9%
23GLM 4.7· Epoch AI Benchmarking Hub data10.9%
24Kimi K2.5benchmark-owner leaderboard run · site model-mode label: Kimi K2.5 · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs9.6%
25DeepSeek V3.2benchmark-owner leaderboard run · site model-mode label: DeepSeek V3.2 Thinking · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.4 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs9.5%
26DeepSeek V3.2 Expthinking· Epoch AI Benchmarking Hub data9.5%
27DeepSeek V3.2· Epoch AI Benchmarking Hub data7.4%
28DeepSeek V3.2benchmark-owner leaderboard run · site model-mode label: DeepSeek V3.2 · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs7.4%
29MiMo V2 Pro· Epoch AI Benchmarking Hub data6.9%
30MiMo V2 Probenchmark-owner leaderboard run · site model-mode label: MiMo v2 Pro (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs6.9%
31MiniMax M2.5· Epoch AI Benchmarking Hub data6.3%
32MiniMax M2.5benchmark-owner leaderboard run · site model-mode label: MiniMax M2.5 (Thinking) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±1 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs6.3%

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 (CL-bench label)benchmark-owner leaderboard run · site model-mode label: DeepSeek V4 Pro (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±1.4 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs13.5%
GLM-4.7 (CL-bench label)benchmark-owner leaderboard run · site model-mode label: GLM-4.7 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs10.9%
GPT-5.1 (CL-bench label)benchmark-owner leaderboard run · site model-mode label: GPT 5.1 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs17.3%
Grok 4.20benchmark-owner leaderboard run · site model-mode label: Grok 4.20 · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs8.6%
Grok 4.20benchmark-owner leaderboard run · site model-mode label: Grok 4.20 (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.4 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs11.9%
Hy2.0 (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Hy2.0 Thinking · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.7 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.8.4%
Hy3 Preview (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Hy3 preview (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.9 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.15.7%
Hy3 Preview (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Hy3 preview (Low) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.12.8%
Hy3 Preview (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Hy3 preview (NoThink) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.9.4%
Qwen3.5 Plusbenchmark-owner leaderboard run · site model-mode label: Qwen 3.5 Plus · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs8.1%
Qwen3.5 Plusbenchmark-owner leaderboard run · site model-mode label: Qwen 3.5 Plus (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±0.1 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs12.4%
Seed 2.0 Pro (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Seed 2.0 Pro · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs8.4%
Seed 2.0 Pro (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Seed 2.0 Pro (High) · 405 tasks · GPT-5.1 judge · high reasoning effort · site standard deviation ±1.7 percentage points · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs15.5%
Seed 2.0 Pro (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Seed 2.0 Pro (Low) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs12.1%
Seed 2.0 Pro (CL-bench label)benchmark-owner leaderboard run · site model-mode label: Seed 2.0 Pro (Medium) · 405 tasks · GPT-5.1 judge · high reasoning effort · single run; no standard deviation · candidate API snapshot and exact candidate harness not specified on public leaderboard· Reported by Tencent Hunyuan / Fudan NLP Group· via BenchBoard docs11.4%

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. · Epoch AI Benchmarking Hub, CC BY 4.0 (licence)