benchgap
Agentic · terminal

Terminal-Bench 3.0 leaderboard

As of 2026-11-18, the highest measured score on Terminal-Bench 3.0 is 42.7% by Claude Opus 5. 104 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.250.4%estimated ± 3.2 pp, medium confidence
2Claude Opus 5.550.4%estimated ± 3.2 pp, medium confidence
3Gemini 4 Carbon49.6%estimated ± 3.2 pp, medium confidence
4Claude Sonnet 5.549.1%estimated ± 3.2 pp, medium confidence
5GPT-6.1 Sol47.1%estimated ± 3.2 pp, medium confidence
6GPT-6 Astra46.8%estimated ± 3.2 pp, medium confidence
7Kimi K3.1 Pro46.4%estimated ± 3.2 pp, medium confidence
8Claude Fable 5.146.0%estimated ± 3.0 pp, medium confidence
9Gemini 4 Hydrogen44.2%estimated ± 3.2 pp, medium confidence
10Gemini 4 Argon43.6%estimated ± 3.2 pp, medium confidence
11Claude Opus 542.7%measured
12GPT-6.1 Terra42.1%estimated ± 3.2 pp, high confidence
13GPT-6 Sol40.2%estimated ± 3.2 pp, high confidence
14Muse Wonder38.0%estimated ± 3.2 pp, high confidence
15SWE-236.4%estimated ± 9.1 pp, low confidence
16GPT-5.6 Sol34.6%measured
17Claude Fable 534.1%measured
18Claude Haiku 5.533.4%estimated ± 3.2 pp, high confidence
19Grok 4.732.6%estimated ± 3.2 pp, high confidence
20DeepSeek V4.1 Flash 101732.4%estimated ± 3.2 pp, high confidence
21Hy432.0%estimated ± 3.2 pp, high confidence
22GLM-5.331.6%estimated ± 3.0 pp, high confidence
23GLM-5.3-Flash29.5%estimated ± 3.0 pp, high confidence
24Muse Ember27.2%estimated ± 3.2 pp, high confidence
25Grok 4.626.5%measured
26SWE-2 Flash25.4%estimated ± 9.1 pp, low confidence
27Mistral Large 425.2%estimated ± 3.2 pp, high confidence
28MiMo-V2.6-Pro24.8%estimated ± 3.2 pp, high confidence
29Qwen3.8-Flash-Next24.5%estimated ± 8.6 pp, low confidence
30LongCat-2.523.7%estimated ± 9.1 pp, low confidence
31Claude Mythos 523.0%estimated ± 9.1 pp, low confidence
32Kimi K322.2%estimated ± 3.0 pp, high confidence
33Ember-121.4%estimated ± 3.2 pp, high confidence
34Gemini 3.8 Flash21.4%estimated ± 3.0 pp, high confidence
35Claude Opus 4.821.1%measured
36GPT-5.521.0%estimated ± 6.9 pp, medium confidence
37GPT-5.6 Terra20.8%measured
38Inkling 2 Small18.5%estimated ± 8.6 pp, low confidence
39Step 5 Preview18.3%estimated ± 9.1 pp, low confidence
40Gemini 3.5 Flash18.2%estimated ± 6.9 pp, medium confidence
41MAI Code 217.8%estimated ± 9.1 pp, low confidence
42MiMo-V2.6-Flash17.7%estimated ± 3.2 pp, high confidence
43Gemini 3.6 Flash17.7%estimated ± 6.9 pp, medium confidence
44GPT-6 Luna17.1%estimated ± 3.2 pp, high confidence
45Muse Spark 1.316.9%estimated ± 3.2 pp, high confidence
46Ultramarine 1 Flash16.3%estimated ± 9.1 pp, low confidence
47Grok 4.515.7%measured
48Sakana Fugu-Ultra15.2%estimated ± 9.1 pp, low confidence
49Gemini 3.7 Flash14.9%measured
50SWE-1.714.6%estimated ± 9.1 pp, low confidence
51Claude Sonnet 514.6%measured
52Ling 3.1 Flash14.4%estimated ± 9.1 pp, low confidence
53GPT-5.6 Luna14.3%measured
54Ripple-65B Preview13.8%estimated ± 9.1 pp, low confidence
55Sakana Fugu13.6%estimated ± 9.1 pp, low confidence
56Ornith-1.5-397B13.5%measured
57Particle-113.3%estimated ± 9.1 pp, low confidence
58Muse Spark 1.213.1%estimated ± 6.9 pp, medium confidence
59Muse Spark 1.112.7%estimated ± 6.9 pp, medium confidence
60Beam 1.0 501B-A23B FP812.6%estimated ± 9.1 pp, low confidence
61Beam 1.0 501B-A23B NVFP412.3%estimated ± 9.1 pp, low confidence
62Atria Dawn Preview12.2%estimated ± 9.1 pp, low confidence
63MAI Code 2 Flash12.2%estimated ± 9.1 pp, low confidence
64MiniMax M312.1%estimated ± 3.2 pp, medium confidence
65Claude Opus 4.711.9%estimated ± 6.9 pp, medium confidence
66Ornith-1.0-397B11.7%estimated ± 9.1 pp, low confidence
67DeepSeek V4 Pro11.5%estimated ± 3.2 pp, medium confidence
68Beam 1.0 501B-A23B11.3%estimated ± 3.2 pp, medium confidence
69Qwen3.8 Max10.9%estimated ± 6.9 pp, low confidence
70DeepSeek V4 Flash 073110.6%estimated ± 6.9 pp, low confidence
71Kimi K2.7 Code10.6%estimated ± 6.9 pp, low confidence
72dots3-note Preview10.3%estimated ± 9.1 pp, low confidence
73Jais-3-375B-A23B9.7%estimated ± 9.1 pp, low confidence
74OrcaCyber-Nano-1.09.1%estimated ± 9.1 pp, low confidence
75Gemini 3.1 Pro8.9%estimated ± 3.2 pp, medium confidence
76Seed 2.1 Pro8.5%estimated ± 9.1 pp, low confidence
77Apodex 1.18.4%estimated ± 9.1 pp, low confidence
78GPT-5.4 mini8.3%estimated ± 3.2 pp, medium confidence
79Laguna S 2.18.2%estimated ± 9.1 pp, low confidence
80Quasar 438B7.8%estimated ± 9.1 pp, low confidence
81Inkling-Small7.7%estimated ± 3.2 pp, medium confidence
82Seed 2.1 Turbo7.3%estimated ± 9.1 pp, low confidence
83Nemotron 3.5 Super VL 120B A12B7.3%estimated ± 3.2 pp, medium confidence
84Jais-3-MoVA-36B-A4B7.2%estimated ± 9.1 pp, low confidence
85Inkling7.1%estimated ± 3.2 pp, medium confidence
86Pokee-Isaac 28B6.5%estimated ± 9.1 pp, low confidence
87Ornith-1.0-35B6.3%estimated ± 9.1 pp, low confidence
88Qwen3.7 Max6.3%estimated ± 6.9 pp, low confidence
89Jais-3-32B6.2%estimated ± 9.1 pp, low confidence
90MiMo-V2.56.1%estimated ± 6.9 pp, low confidence
91MAI-Code-1.1-Flash6.0%estimated ± 9.1 pp, low confidence
92Nemotron 3.5 Super VL 120B A12B NVFP45.5%estimated ± 9.1 pp, low confidence
93Step 3.7 Flash5.2%estimated ± 9.1 pp, low confidence
94Ornith-1.5-35B-A3B5.1%measured
95Composer 2.54.9%estimated ± 6.9 pp, low confidence
96Qwen3.8-27B4.9%estimated ± 6.9 pp, low confidence
97Solar Pro 44.7%estimated ± 9.1 pp, low confidence
98GLM-5.24.6%measured
99Claude Sonnet 4.64.4%estimated ± 6.9 pp, low confidence
100MiMo-V2.5-Pro4.4%estimated ± 6.9 pp, low confidence
101Flint4.3%estimated ± 9.1 pp, low confidence
102GLM-5.14.2%estimated ± 6.9 pp, low confidence
103Nemotron 3.5 Bolt 74B-A9B BF164.2%estimated ± 9.1 pp, low confidence
104Ternary Bonsai 2 27B4.0%estimated ± 9.1 pp, low confidence
105Muse Glimmer 30B3.8%estimated ± 9.1 pp, low confidence
106Nemotron 3.5 Bolt 74B-A9B3.7%estimated ± 9.1 pp, low confidence
107Hy4 preview3.5%estimated ± 6.9 pp, low confidence
108Jais-3-7B3.1%estimated ± 9.1 pp, low confidence
109Ornith-1.5-9B3.1%estimated ± 9.1 pp, low confidence
110Gemini 3 Flash3.1%estimated ± 6.9 pp, low confidence
111Kimi K2.63.0%estimated ± 6.9 pp, low confidence
112Qwen3.6 Plus2.8%estimated ± 6.9 pp, low confidence
113K-EXAONE 2.02.8%estimated ± 9.1 pp, low confidence
114Qwen3.7 Plus2.7%estimated ± 6.9 pp, low confidence
115Nemotron 3 Ultra2.2%estimated ± 6.9 pp, low confidence
116Gemini 3.5 Flash-Lite2.0%estimated ± 6.9 pp, low confidence
117Ling 3.0 Flash2.0%estimated ± 6.9 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General