benchgap
Agentic · terminal

AA Terminal-Bench 2.1 leaderboard

As of 2026-11-21, the highest measured score on AA Terminal-Bench 2.1 is 91.4% by Claude Fable 5.1. 88 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.191.4%measured
2Claude Opus 5.590.0%estimated ± 2.5 pp, medium confidence
3GPT-6 Astra89.9%estimated ± 2.5 pp, medium confidence
4Claude Opus 589.1%measured
5GPT-5.6 Sol89.1%estimated ± 2.5 pp, medium confidence
6Grok 4.688.4%measured
7GPT-5.6 Terra88.0%measured
8Claude Sonnet 5.587.7%estimated ± 2.5 pp, high confidence
9GPT-6 Sol87.7%estimated ± 2.5 pp, high confidence
10Gemini 3.8 Flash87.6%measured
11SWE-286.2%estimated ± 2.6 pp, medium confidence
12SWE-2 Flash86.2%estimated ± 2.6 pp, high confidence
13LongCat-2.586.2%estimated ± 2.6 pp, high confidence
14Claude Mythos 586.2%estimated ± 2.6 pp, high confidence
15Qwen3.8-Flash-Next86.1%measured
16Gemini 3.7 Flash85.8%measured
17Ornith-1.5-397B85.2%estimated ± 2.6 pp, high confidence
18Kimi K385.0%measured
19MiMo-V2.6-Flash84.8%estimated ± 2.5 pp, high confidence
20Claude Fable 584.6%measured
21Claude Opus 4.884.6%measured
22GLM-5.3-Flash84.3%measured
23GPT-5.584.3%measured
24DeepSeek V4.1 Flash 101784.1%estimated ± 2.5 pp, high confidence
25Gemini 3.5 Flash84.0%estimated ± 2.5 pp, high confidence
26GLM-5.383.9%measured
27Gemini 3.6 Flash83.9%estimated ± 2.5 pp, high confidence
28Grok 4.783.7%estimated ± 2.5 pp, high confidence
29GPT-6 Luna83.6%estimated ± 2.5 pp, high confidence
30Muse Spark 1.383.4%estimated ± 2.5 pp, high confidence
31Claude Opus 4.783.1%measured
32Step 5 Preview83.0%estimated ± 2.6 pp, high confidence
33Gemini 3.1 Pro82.9%estimated ± 2.5 pp, high confidence
34Muse Spark 1.182.5%estimated ± 2.5 pp, high confidence
35MAI Code 282.4%estimated ± 2.6 pp, high confidence
36GLM-5.282.1%estimated ± 2.5 pp, high confidence
37MiMo-V2.6-Pro82.1%estimated ± 2.5 pp, high confidence
38Qwen3.8 Max82.0%estimated ± 2.5 pp, high confidence
39DeepSeek V4 Flash 073181.9%estimated ± 2.5 pp, high confidence
40Kimi K2.7 Code81.9%estimated ± 2.5 pp, high confidence
41Ultramarine 1 Flash81.8%estimated ± 2.6 pp, high confidence
42Sakana Fugu-Ultra81.7%estimated ± 2.6 pp, high confidence
43Ember-181.7%estimated ± 2.6 pp, high confidence
44SWE-1.781.7%estimated ± 2.6 pp, high confidence
45Ling 3.1 Flash81.7%estimated ± 2.6 pp, high confidence
46Ripple-65B Preview81.7%estimated ± 2.6 pp, high confidence
47Sakana Fugu81.7%estimated ± 2.6 pp, high confidence
48Beam 1.0 501B-A23B81.7%estimated ± 2.6 pp, high confidence
49Particle-181.7%estimated ± 2.6 pp, high confidence
50Beam 1.0 501B-A23B FP881.7%estimated ± 2.6 pp, high confidence
51Beam 1.0 501B-A23B NVFP481.7%estimated ± 2.6 pp, high confidence
52Atria Dawn Preview81.7%estimated ± 2.6 pp, high confidence
53MAI Code 2 Flash81.7%estimated ± 2.6 pp, high confidence
54Ornith-1.0-397B81.7%estimated ± 2.6 pp, high confidence
55dots3-note Preview81.7%estimated ± 2.6 pp, high confidence
56Jais-3-375B-A23B81.7%estimated ± 2.6 pp, high confidence
57OrcaCyber-Nano-1.081.7%estimated ± 2.6 pp, medium confidence
58Inkling 2 Mini81.7%estimated ± 2.6 pp, medium confidence
59Seed 2.1 Pro81.7%estimated ± 2.6 pp, medium confidence
60Apodex 1.181.7%estimated ± 2.6 pp, medium confidence
61Laguna S 2.181.7%estimated ± 2.6 pp, medium confidence
62Quasar 438B81.7%estimated ± 2.6 pp, medium confidence
63Jais-3-32B81.7%estimated ± 2.6 pp, medium confidence
64Jais-3-MoVA-36B-A4B81.7%estimated ± 2.6 pp, medium confidence
65MAI-Code-1.1-Flash81.7%estimated ± 2.6 pp, medium confidence
66Nemotron 3.5 Super VL 120B A12B81.7%estimated ± 2.6 pp, medium confidence
67Nemotron 3.5 Super VL 120B A12B NVFP481.7%estimated ± 2.6 pp, medium confidence
68Ornith-1.0-35B81.7%estimated ± 2.6 pp, medium confidence
69Ornith-1.5-35B-A3B81.7%estimated ± 2.6 pp, medium confidence
70Pokee-Isaac 28B81.7%estimated ± 2.6 pp, medium confidence
71Seed 2.1 Turbo81.7%estimated ± 2.6 pp, medium confidence
72Solar Pro 481.7%estimated ± 2.6 pp, medium confidence
73Step 3.7 Flash81.7%estimated ± 2.6 pp, medium confidence
74Grok 4.581.6%measured
75Inkling 2 Small81.3%measured
76GPT-5.6 Luna80.9%measured
77Qwen3.7 Max80.8%estimated ± 2.5 pp, high confidence
78MiMo-V2.580.8%estimated ± 2.5 pp, high confidence
79Claude Sonnet 580.5%measured
80Composer 2.580.4%estimated ± 2.5 pp, high confidence
81Gemini 4 Argon80.4%estimated ± 2.5 pp, medium confidence
82Claude Sonnet 4.680.3%estimated ± 2.5 pp, medium confidence
83MiMo-V2.5-Pro80.3%estimated ± 2.5 pp, medium confidence
84GLM-5.180.3%estimated ± 2.5 pp, medium confidence
85Muse Spark 1.280.1%measured
86Hy4 preview80.1%estimated ± 2.5 pp, medium confidence
87Inkling-Small80.1%estimated ± 2.5 pp, medium confidence
88DeepSeek V4 Pro80.0%estimated ± 2.5 pp, medium confidence
89GPT-5.4 mini80.0%estimated ± 2.5 pp, medium confidence
90Gemini 3 Flash80.0%estimated ± 2.5 pp, medium confidence
91Kimi K2.680.0%estimated ± 2.5 pp, medium confidence
92MiniMax M380.0%estimated ± 2.5 pp, medium confidence
93Qwen3.6 Plus79.9%estimated ± 2.5 pp, medium confidence
94Qwen3.7 Plus79.9%estimated ± 2.5 pp, medium confidence
95Qwen3.8-27B79.8%measured
96Nemotron 3 Ultra79.8%estimated ± 2.5 pp, medium confidence
97Gemini 3.5 Flash-Lite79.7%estimated ± 2.5 pp, medium confidence
98Ling 3.0 Flash79.7%estimated ± 2.5 pp, medium confidence
99MiniMax M2.779.6%estimated ± 2.5 pp, medium confidence
100Inkling79.6%estimated ± 2.5 pp, medium confidence
101Grok 4.2079.4%estimated ± 2.5 pp, medium confidence
102Claude Haiku 4.579.4%estimated ± 2.5 pp, medium confidence
103Grok 4.379.4%estimated ± 2.5 pp, medium confidence
104GPT-5.4 nano79.4%estimated ± 2.5 pp, medium confidence
105Mistral Medium 3.5 128B79.3%estimated ± 2.5 pp, medium confidence
106Mercury 2.579.2%estimated ± 2.5 pp, medium confidence
107Gemini 3.1 Flash-Lite79.2%estimated ± 2.5 pp, medium confidence
108Laguna M.179.2%estimated ± 2.5 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General