Skip to content

Benchmark

CHI-Bench

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
23
Scores recorded
40
Sources
2
IQ weight
2.34

Score distribution

16.8%
Mean
15.6%
Median
10.9%
Std dev
4.4%
10th percentile
28%
90th percentile
0%
Min
0%54.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.34 for chat models, measured over 23 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Claude Opus 5clinical-result-135· Clinical Benchmarks data docs54.7%
2Claude Opus 4.8clinical-result-136· Clinical Benchmarks data docs37.3%
3Claude Opus 5clinical-result-137· Clinical Benchmarks data docs37.3%
4Claude Opus 4.8clinical-result-138· Clinical Benchmarks data docs33.3%
5Claude Opus 4.6clinical-result-139· Clinical Benchmarks data docs28%
6Claude Sonnet 4.6clinical-result-140· Clinical Benchmarks data docs26.2%
7GPT-5.6 Solclinical-result-141· Clinical Benchmarks data docs25.3%
8Kimi K3clinical-result-142· Clinical Benchmarks data docs25.3%
9Claude Opus 4.7clinical-result-143· Clinical Benchmarks data docs24.4%
10Claude Fable 5clinical-result-144· Clinical Benchmarks data docs24%
11GPT 5.5clinical-result-326· Clinical Benchmarks data docs20.9%
12Claude Sonnet 5clinical-result-327· Clinical Benchmarks data docs20%
13GLM 5.1clinical-result-2293· Clinical Benchmarks data docs18.7%
14GLM 5.1clinical-result-2294· Clinical Benchmarks data docs18.7%
15GLM 5.2clinical-result-2295· Clinical Benchmarks data docs18.7%
16Claude Opus 4.7clinical-result-2296· Clinical Benchmarks data docs17.3%
17GLM 5.1clinical-result-2297· Clinical Benchmarks data docs16.9%
18GPT 5.4clinical-result-2299· Clinical Benchmarks data docs16%
19Kimi K2.6· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim15.6%
20Kimi K2.6clinical-result-2301· Clinical Benchmarks data docs15.6%
21Kimi K2.6clinical-result-2302· Clinical Benchmarks data docs15.1%
22DeepSeek V4 Pro· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim14.2%
23DeepSeek V4 Proclinical-result-2303· Clinical Benchmarks data docs14.2%
24DeepSeek V4 Proclinical-result-2304· Clinical Benchmarks data docs13.8%
25GPT-5.6 Lunaclinical-result-2306· Clinical Benchmarks data docs13.3%
26GPT-5.6 Terraclinical-result-2305· Clinical Benchmarks data docs13.3%
27Gemini 3 Flashclinical-result-2307· Clinical Benchmarks data docs12.5%
28DeepSeek V4 Proclinical-result-2308· Clinical Benchmarks data docs11.1%
29GLM 5.1clinical-result-2309· Clinical Benchmarks data docs11.1%
30DeepSeek V4 Proclinical-result-2310· Clinical Benchmarks data docs10.7%
31Kimi K2.6clinical-result-2311· Clinical Benchmarks data docs10.2%
32GPT-5.4 miniclinical-result-2313· Clinical Benchmarks data docs8.4%
33Inkling FP4clinical-result-2314· Clinical Benchmarks data docs8%
34Claude Haiku 4.5clinical-result-2315· Clinical Benchmarks data docs6.2%
35Grok 4.3clinical-result-2316· Clinical Benchmarks data docs5.8%
36Grok 4.3clinical-result-2318· Clinical Benchmarks data docs4.4%
37Kimi K2.6clinical-result-2319· Clinical Benchmarks data docs3.1%
38Grok 4.3clinical-result-2320· Clinical Benchmarks data docs2.2%
39Grok 4.3clinical-result-2321· Clinical Benchmarks data docs0.4%
40Nemotron 3 Ultraclinical-result-2322· Clinical Benchmarks data docs0%

Data credits: Clinical Benchmarks (clinicalbenchmarks.ai), CC BY 4.0; fractions converted to percentages; source documents retain their own terms (licence) · Hugging Face official benchmark leaderboards; licence varies by benchmark