benchgap
Agentic · terminal

Terminal-Bench 2.1 (Vals AI) leaderboard

As of 2026-11-21, the highest measured score on Terminal-Bench 2.1 (Vals AI) is 87.6% by Claude Opus 5.5. 99 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Opus 5.587.6%measured
2GPT-6 Astra87.3%measured
3GPT-5.6 Sol85.8%measured
4Kimi K3.1 Pro85.3%estimated ± 6.9 pp, low confidence
5Claude Fable 5.185.0%measured
6Gemini 4 Carbon84.6%estimated ± 6.9 pp, low confidence
7Claude Opus 584.6%measured
8GPT-6.2 Astra84.2%estimated ± 6.9 pp, low confidence
9Gemini 4 Neon84.0%estimated ± 6.9 pp, low confidence
10Claude Sonnet 5.583.1%measured
11GPT-6 Sol83.1%measured
12GPT-6.2 Sol82.7%estimated ± 6.9 pp, low confidence
13Claude Fable 5.282.6%estimated ± 6.9 pp, low confidence
14Muse Wonder Full81.7%estimated ± 6.9 pp, low confidence
15Gemini 3.8 Flash81.3%measured
16GPT-6.1 Terra81.1%estimated ± 6.9 pp, low confidence
17Kimi K380.9%measured
18Claude Fable 580.5%measured
19Gemini 4 Hydrogen80.3%estimated ± 6.9 pp, low confidence
20GPT-6.1 Sol79.9%estimated ± 8.3 pp, medium confidence
21Muse Wonder79.8%estimated ± 6.9 pp, low confidence
22GPT-5.6 Luna79.0%measured
23Qwen4 Max78.5%estimated ± 6.9 pp, low confidence
24Grok 4.678.3%measured
25Hy478.0%estimated ± 8.3 pp, medium confidence
26Gemini 3.7 Flash77.5%measured
27GPT-5.6 Terra77.5%measured
28SWE-277.3%estimated ± 7.3 pp, low confidence
29Qwen3.8-Flash-Next76.9%estimated ± 5.5 pp, low confidence
30GPT-5.576.4%measured
31MiMo-V2.6-Flash76.4%measured
32Muse Ember75.8%estimated ± 6.9 pp, low confidence
33Ornith-1.5-397B75.0%estimated ± 3.4 pp, high confidence
34SWE-2 Flash74.6%estimated ± 7.3 pp, medium confidence
35Claude Sonnet 574.5%measured
36DeepSeek V4.1 Flash 101774.5%measured
37Gemini 3.5 Flash74.2%measured
38LongCat-2.574.0%estimated ± 7.3 pp, medium confidence
39Gemini 3.6 Flash73.8%measured
40Claude Mythos 573.7%estimated ± 7.3 pp, medium confidence
41Grok 4.773.4%measured
42GPT-6 Luna73.0%measured
43Claude Haiku 5.572.8%estimated ± 6.9 pp, low confidence
44Muse Spark 1.372.3%measured
45Claude Opus 4.871.9%measured
46GPT-5.471.6%estimated ± 5.6 pp, medium confidence
47GLM-5.371.5%measured
48Step 5 Preview71.4%estimated ± 7.3 pp, medium confidence
49MAI Code 271.1%estimated ± 7.3 pp, medium confidence
50Gemini 3.1 Pro70.8%measured
51Hy4 flash70.4%estimated ± 8.3 pp, medium confidence
52Ultramarine 1 Flash70.0%estimated ± 7.3 pp, medium confidence
53Mistral Large 469.8%estimated ± 6.9 pp, low confidence
54Inkling 2 Small69.7%estimated ± 5.5 pp, low confidence
55Muse Spark 1.269.7%measured
56Muse Spark 1.169.3%measured
57Sakana Fugu-Ultra69.2%estimated ± 7.3 pp, medium confidence
58Ember-169.1%estimated ± 7.3 pp, medium confidence
59GPT-5.3 Codex68.8%estimated ± 6.4 pp, medium confidence
60SWE-1.768.7%estimated ± 7.3 pp, medium confidence
61Claude Opus 4.768.5%measured
62Ling 3.1 Flash68.5%estimated ± 7.3 pp, medium confidence
63Ripple-65B Preview68.0%estimated ± 7.3 pp, medium confidence
64GLM-5.267.8%measured
65Grok 4.567.8%measured
66MiMo-V2.6-Pro67.8%measured
67Sakana Fugu67.7%estimated ± 7.3 pp, medium confidence
68Particle-167.5%estimated ± 7.3 pp, medium confidence
69Qwen3.8 Max67.4%measured
70DeepSeek V4 Flash 073167.0%measured
71Kimi K2.7 Code67.0%measured
72Beam 1.0 501B-A23B FP866.8%estimated ± 7.3 pp, medium confidence
73Beam 1.0 501B-A23B NVFP466.4%estimated ± 7.3 pp, medium confidence
74Atria Dawn Preview66.3%estimated ± 7.3 pp, medium confidence
75MAI Code 2 Flash66.3%estimated ± 7.3 pp, medium confidence
76Ornith-1.0-397B65.7%estimated ± 7.3 pp, medium confidence
77dots3-note Preview63.9%estimated ± 7.3 pp, medium confidence
78Jais-3-375B-A23B62.9%estimated ± 7.3 pp, medium confidence
79GLM-5.3-Flash62.9%measured
80Beam 1.0 501B-A23B61.9%estimated ± 6.9 pp, low confidence
81OrcaCyber-Nano-1.061.9%estimated ± 7.3 pp, medium confidence
82Ornith-1.5-35B-A3B61.7%estimated ± 3.4 pp, medium confidence
83Inkling 2 Mini61.2%estimated ± 7.3 pp, medium confidence
84Qwen3.7 Max61.0%measured
85Seed 2.1 Pro60.8%estimated ± 7.3 pp, medium confidence
86MiMo-V2.560.7%measured
87Apodex 1.160.6%estimated ± 7.3 pp, medium confidence
88Laguna S 2.160.2%estimated ± 7.3 pp, medium confidence
89Claude Opus 4.660.0%estimated ± 5.6 pp, medium confidence
90Quasar 438B59.5%estimated ± 7.3 pp, medium confidence
91Claude Opus 4.5 Thinking58.6%estimated ± 5.6 pp, medium confidence
92GPT-5.258.6%estimated ± 5.6 pp, medium confidence
93Composer 2.558.4%measured
94Qwen3.8-27B58.4%measured
95Seed 2.1 Turbo58.2%estimated ± 7.3 pp, medium confidence
96Jais-3-MoVA-36B-A4B58.0%estimated ± 7.3 pp, medium confidence
97Gemini 4 Argon57.7%measured
98Nemotron 3.5 Super VL 120B A12B57.5%estimated ± 7.3 pp, medium confidence
99GPT-5.157.4%estimated ± 5.6 pp, medium confidence
100Muse Spark57.4%estimated ± 5.6 pp, medium confidence
101Claude Sonnet 4.657.3%measured
102MiMo-V2.5-Pro57.3%measured
103GLM-5.156.9%measured
104Pokee-Isaac 28B56.3%estimated ± 7.3 pp, medium confidence
105Qwen 3.6 Max (preview)56.2%estimated ± 5.6 pp, medium confidence
106GLM-555.7%estimated ± 5.6 pp, medium confidence
107Ornith-1.0-35B55.6%estimated ± 7.3 pp, medium confidence
108Jais-3-32B55.3%estimated ± 7.3 pp, medium confidence
109Hy4 preview55.1%measured
110Inkling-Small55.1%measured
111DeepSeek V4 Pro54.7%measured
112GPT-5.4 mini54.7%measured
113MAI-Code-1.1-Flash54.7%estimated ± 7.3 pp, medium confidence
114Composer 254.5%estimated ± 6.4 pp, medium confidence
115Gemini 3 Flash53.9%measured
116Kimi K2.653.6%measured
117MiniMax M353.6%measured
118Qwen3.6 Plus53.2%measured
119Nemotron 3.5 Super VL 120B A12B NVFP453.1%estimated ± 7.3 pp, medium confidence
120Qwen3.7 Plus52.8%measured
121Claude Opus 4.552.3%estimated ± 6.4 pp, medium confidence
122Qwen3.6-27B52.3%estimated ± 6.4 pp, medium confidence
123Step 3.7 Flash52.1%estimated ± 7.3 pp, medium confidence
124Solar Open 251.2%estimated ± 5.6 pp, low confidence
125Nemotron 3 Ultra50.9%measured
126Gemini 3.5 Flash-Lite50.2%measured
127Ling 3.0 Flash50.2%measured
128Solar Pro 450.2%estimated ± 7.3 pp, medium confidence
129MiniMax M2.748.7%measured
130Flint48.6%estimated ± 7.3 pp, medium confidence
131Nemotron 3.5 Bolt 74B-A9B BF1647.9%estimated ± 7.3 pp, low confidence
132Hy3 Preview47.8%estimated ± 6.4 pp, medium confidence
133Inkling47.6%measured
134Ternary Bonsai 2 27B47.0%estimated ± 7.3 pp, low confidence
135Muse Glimmer 30B46.2%estimated ± 7.3 pp, low confidence
136Qwen3.5 397B46.1%estimated ± 6.4 pp, medium confidence
137Nemotron 3.5 Bolt 74B-A9B45.6%estimated ± 7.3 pp, low confidence
138Qwen3.6-35B-A3B45.2%estimated ± 6.4 pp, medium confidence
139Kimi K2.544.5%estimated ± 6.4 pp, medium confidence
140Grok 4.2044.2%measured
141Claude Haiku 4.543.8%measured
142Claude Sonnet 4.543.8%estimated ± 6.4 pp, medium confidence
143Qwen3.5-122B-A10B43.2%estimated ± 6.4 pp, medium confidence
144Jais-3-7B42.2%estimated ± 7.3 pp, low confidence
145Ornith-1.5-9B42.0%estimated ± 7.3 pp, low confidence
146Grok 4.341.9%measured
147GPT-5.4 nano41.6%measured
148K-EXAONE 2.040.2%estimated ± 7.3 pp, low confidence
149MAI-Thinking-140.1%estimated ± 6.4 pp, medium confidence
150Ornith-1.0-9B39.7%estimated ± 7.3 pp, low confidence
151Mistral Medium 3.5 128B39.0%measured
152Jais-3-3.7B38.2%estimated ± 7.3 pp, low confidence
153Qwen3.5-27B36.1%estimated ± 6.4 pp, medium confidence
154GLM-4.735.5%estimated ± 6.4 pp, medium confidence
155Qwen3.5-35B-A3B35.1%estimated ± 6.4 pp, medium confidence
156Mercury 2.534.5%measured
157A.X K234.3%estimated ± 7.3 pp, low confidence
158Gemini 3.1 Flash-Lite34.1%measured
159Laguna M.134.1%measured
160Laguna XS 2.132.3%estimated ± 6.4 pp, medium confidence
161Granite 4.2 30B29.2%estimated ± 7.3 pp, low confidence
162LongCat-Flash-Lite-Sparse28.8%estimated ± 6.4 pp, low confidence
163Laguna XS.225.8%measured
164Nemotron 3.5 Lightning 30B A3B NVFP424.8%estimated ± 7.3 pp, low confidence
165Granite 4.2 8B22.6%estimated ± 7.3 pp, low confidence
166Mellum2.1-12B-A2.5B-Thinking20.2%estimated ± 7.3 pp, low confidence
167Command A+17.6%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General