Agentic · tools

AA-AnalystAgent leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on AA-AnalystAgent is 60.0% by Gemini 3.7 Flash (high). 28 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: artificial-analysis.

#	Model	Score	Source
1	Gemini 3.7 Flash (high)	60.0%	measured
2	Gemini 4 Argon (high)	58.3%	estimated ± 7.8 pp, low confidence
3	Claude Fable 5.1 (max with fallback)	57.5%	measured
4	Claude Sonnet 5.5 (max with fallback)	57.5%	measured
5	Claude Opus 5.5 (max with fallback)	56.3%	measured
6	Qwen3.8 Max	54.8%	estimated ± 8.1 pp, low confidence
7	DeepSeek V4.1 Flash (max)	54.8%	estimated ± 7.8 pp, medium confidence
8	Claude Opus 5 (max)	53.8%	measured
9	GPT-6 Astra (xhigh)	53.8%	estimated ± 7.8 pp, medium confidence
10	Grok 4.6 (xhigh)	53.6%	estimated ± 7.8 pp, medium confidence
11	GPT-6 Astra (high)	53.4%	estimated ± 7.8 pp, medium confidence
12	GPT-6.1 Sol (xhigh)	53.4%	estimated ± 7.8 pp, medium confidence
13	Grok 4.7 (xhigh)	52.7%	estimated ± 7.8 pp, medium confidence
14	Qwen3.8 2.4T A95B	52.0%	estimated ± 8.1 pp, low confidence
15	GPT-6 Astra (max)	51.2%	measured
16	GPT-6.1 Sol (max)	50.0%	measured
17	GPT-5.5 (xhigh)	50.0%	measured
18	Qwen3.8 27B (medium)	49.8%	estimated ± 8.1 pp, low confidence
19	Muse Spark 1.3 (xhigh)	49.5%	estimated ± 8.1 pp, low confidence
20	GPT-6 Sol (max)	49.4%	estimated ± 7.8 pp, medium confidence
21	Claude Fable 5 (with fallback)	48.8%	measured
22	Gemini 3.8 Flash (high)	47.8%	estimated ± 7.8 pp, medium confidence
23	Mistral Large 4 Preview	47.8%	estimated ± 7.8 pp, medium confidence
24	GPT-5.6 Sol (max)	47.5%	measured
25	Gemini 3.7 Flash (medium)	47.5%	estimated ± 3.6 pp, medium confidence
26	Claude Fable 5.1 (xhigh with fallback)	47.3%	estimated ± 13.9 pp, low confidence
27	Gemini 3.5 Flash (high)	46.8%	estimated ± 3.6 pp, medium confidence
28	MiMo-V2.6-Pro	46.4%	estimated ± 7.8 pp, medium confidence
29	Claude Fable 5.1 (high with fallback)	46.1%	estimated ± 13.9 pp, low confidence
30	DeepSeek V4 Pro 0813 (max)	45.7%	estimated ± 3.6 pp, medium confidence
31	Muse Spark 1.3 (max)	45.6%	estimated ± 7.8 pp, medium confidence
32	Qwen3.8 Max (0902)	45.0%	measured
33	Grok 4.6 (high)	43.0%	estimated ± 3.6 pp, medium confidence
34	GPT-6 Luna (max)	39.6%	estimated ± 7.8 pp, medium confidence
35	Kimi K3 (max)	38.8%	measured
36	Step 5 Preview	35.0%	measured
37	Qwen3.8 27B (xhigh)	34.5%	estimated ± 3.6 pp, medium confidence
38	Inkling (xhigh)	23.8%	measured
39	GLM-5.3 (max)	19.4%	estimated ± 3.6 pp, medium confidence
40	K2 Horizon 375B A23B	19.1%	estimated ± 7.8 pp, medium confidence
41	Muse Glimmer (high)	16.2%	estimated ± 3.6 pp, medium confidence
42	GLM-5.3-Flash	13.5%	estimated ± 3.6 pp, medium confidence
43	MiniMax-M3	10.0%	measured
44	Nemotron 3 Ultra	6.3%	measured

AA-AnalystAgent leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Vision & documents