benchgap
Agentic · tools

AA AutomationBench leaderboard

As of 2026-11-21, the highest measured score on AA AutomationBench is 79.3% by Gemini 4 Neon. 65 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Gemini 4 Neon79.3%measured
2Gemini 4 Carbon77.8%measured
3GPT-6 Cyber77.6%estimated ± 4.8 pp, medium confidence
4Gemini 4 Argon77.5%measured
5GPT-6.2 Astra77.2%measured
6Claude Fable 5.276.4%measured
7GPT-6 Astra76.2%estimated ± 4.8 pp, medium confidence
8Gemini 4 Hydrogen76.1%measured
9Claude Mythos 5.275.0%estimated ± 4.8 pp, medium confidence
10GPT-6.1 Sol74.0%estimated ± 4.8 pp, medium confidence
11GPT-5.6 Sol73.4%estimated ± 4.8 pp, medium confidence
12Kimi K3.1 Pro73.3%measured
13GPT-5.6 Cyber72.8%estimated ± 4.8 pp, medium confidence
14Qwen4 Max72.6%measured
15Claude Sonnet 5.571.8%measured
16Claude Fable 571.5%estimated ± 12.8 pp, low confidence
17GPT-6.2 Sol71.3%measured
18Claude Opus 5.569.5%measured
19Claude Mythos 5.168.2%estimated ± 4.8 pp, medium confidence
20Muse Ember68.1%measured
21GPT-5.6 Terra68.0%estimated ± 4.8 pp, medium confidence
22LongCat-2.568.0%estimated ± 7.9 pp, low confidence
23GPT-6 Sol67.2%estimated ± 4.8 pp, medium confidence
24Grok 4.765.6%measured
25Claude Mythos Preview63.2%estimated ± 4.8 pp, medium confidence
26Claude Opus 562.7%estimated ± 5.5 pp, medium confidence
27GLM-5.362.2%measured
28DeepSeek V4.1 Flash 101760.6%estimated ± 4.8 pp, medium confidence
29GPT-6.1 Terra60.2%measured
30Claude Fable 5.159.2%estimated ± 5.5 pp, medium confidence
31MiMo-V2.6-Pro58.6%measured
32GPT-5.558.0%estimated ± 4.8 pp, medium confidence
33Mistral Large 457.8%measured
34GPT-5.6 Luna56.4%estimated ± 4.8 pp, medium confidence
35Gemini 3.1 Pro55.0%estimated ± 12.8 pp, low confidence
36Gemini 3.7 Flash55.0%estimated ± 12.8 pp, low confidence
37Grok 4.655.0%estimated ± 12.8 pp, low confidence
38GLM-5.3-Flash54.6%estimated ± 5.5 pp, medium confidence
39GPT-6 Luna53.2%measured
40Claude Haiku 4.552.8%estimated ± 12.8 pp, low confidence
41Claude Opus 4.852.8%estimated ± 12.8 pp, low confidence
42Claude Sonnet 4.652.8%estimated ± 12.8 pp, low confidence
43Gemini 3.5 Flash52.8%estimated ± 12.8 pp, low confidence
44Gemini 3.6 Flash52.8%estimated ± 12.8 pp, low confidence
45GPT-5.3-Codex-Spark52.8%estimated ± 12.8 pp, low confidence
46Grok 4.2052.8%estimated ± 12.8 pp, low confidence
47Llama 4 Maverick52.8%estimated ± 12.8 pp, low confidence
48MiMo-V2.5-Pro52.8%estimated ± 12.8 pp, low confidence
49Muse Spark 1.252.8%estimated ± 12.8 pp, low confidence
50Muse Spark 1.352.8%estimated ± 12.8 pp, low confidence
51MiniMax M3.1 Flash52.5%estimated ± 11.0 pp, low confidence
52Step 5 Preview51.7%estimated ± 5.5 pp, medium confidence
53Gemini 3.8 Flash51.2%estimated ± 5.5 pp, medium confidence
54Kimi K350.8%estimated ± 5.5 pp, medium confidence
55MiniMax M348.5%estimated ± 8.2 pp, medium confidence
56Beam 1.0 501B-A23B48.5%measured
57Qwen3.8-27B48.2%measured
58Inkling 2 Small45.6%measured
59Muse Wonder Full45.4%measured
60Muse Spark Open44.8%estimated ± 7.9 pp, low confidence
61Ling 3.0 Flash44.2%estimated ± 11.0 pp, low confidence
62DeepSeek V4 Pro42.4%estimated ± 7.9 pp, low confidence
63Hy442.0%estimated ± 7.9 pp, low confidence
64Claude Sonnet 542.0%estimated ± 7.9 pp, low confidence
65Ling 3.0 Flash Sante41.9%estimated ± 7.9 pp, low confidence
66Qwen3.8 Max41.9%estimated ± 7.9 pp, low confidence
67Apodex 1.141.9%estimated ± 7.9 pp, low confidence
68Ornith-1.5-397B41.9%estimated ± 7.9 pp, low confidence
69Hy4 preview41.9%estimated ± 7.9 pp, low confidence
70Hy4 flash41.9%estimated ± 7.9 pp, low confidence
71Sage-A1.541.9%estimated ± 7.9 pp, low confidence
72GLM-5.241.9%estimated ± 7.9 pp, low confidence
73Qwen3.7 Max41.9%estimated ± 7.9 pp, low confidence
74dots3-note Preview41.9%estimated ± 7.9 pp, low confidence
75GLM-5.141.9%estimated ± 7.9 pp, low confidence
76Beam 1.0 501B-A23B FP841.9%estimated ± 7.9 pp, low confidence
77Beam 1.0 501B-A23B NVFP441.9%estimated ± 7.9 pp, low confidence
78GLM-541.9%estimated ± 7.9 pp, low confidence
79Agents-A141.9%estimated ± 7.9 pp, low confidence
80Step 3.7 Flash41.9%estimated ± 7.9 pp, low confidence
81DeepSeek V4 Flash 073141.9%estimated ± 7.9 pp, low confidence
82Nemotron 3 Ultra41.9%estimated ± 7.9 pp, low confidence
83Ornith-1.5-35B-A3B41.9%estimated ± 7.9 pp, low confidence
84Spark-X2.6-4B41.9%estimated ± 7.9 pp, low confidence
85GPT-5.440.6%estimated ± 4.8 pp, low confidence
86MiMo-V2.6-Flash40.6%estimated ± 4.8 pp, low confidence
87Muse Wonder39.1%measured
88Claude Haiku 5.535.4%measured
89Muse Spark 1.18.9%estimated ± 4.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General