benchgap
Agentic · terminal

Terminal-Bench 2.1 leaderboard

As of 2026-11-18, the highest measured score on Terminal-Bench 2.1 is 92.8% by SWE-2. 52 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1SWE-292.8%measured
2GPT-5.6 Sol91.9%measured
3DeepSeek V4.1 Flash 101790.6%measured
4MiMo-V2.6-Pro89.9%measured
5Gemini 3.8 Flash89.4%measured
6SWE-2 Flash89.2%measured
7Kimi K3.1 Pro89.1%estimated ± 5.4 pp, low confidence
8Gemini 4 Carbon89.1%estimated ± 5.4 pp, low confidence
9GPT-6.2 Astra89.1%estimated ± 5.4 pp, low confidence
10Claude Sonnet 5.589.1%estimated ± 5.4 pp, low confidence
11GPT-6.2 Sol89.1%estimated ± 5.4 pp, low confidence
12Claude Fable 5.289.1%estimated ± 5.4 pp, low confidence
13Claude Opus 5.589.1%estimated ± 5.4 pp, low confidence
14Muse Wonder Full89.1%estimated ± 5.4 pp, low confidence
15GPT-6.1 Terra89.1%estimated ± 5.4 pp, low confidence
16Gemini 4 Hydrogen89.1%estimated ± 5.4 pp, low confidence
17Muse Wonder89.1%estimated ± 5.4 pp, low confidence
18Qwen4 Max89.1%estimated ± 5.4 pp, low confidence
19Muse Ember89.1%estimated ± 5.4 pp, low confidence
20Claude Haiku 5.589.1%estimated ± 5.4 pp, low confidence
21Grok 4.789.0%estimated ± 5.4 pp, low confidence
22Mistral Large 489.0%estimated ± 5.4 pp, low confidence
23GPT-6.1 Sol88.9%estimated ± 5.6 pp, low confidence
24GPT-6 Astra88.9%estimated ± 5.6 pp, low confidence
25Claude Fable 5.188.8%estimated ± 5.6 pp, low confidence
26Muse Spark 1.388.8%measured
27Gemini 4 Argon88.7%estimated ± 5.6 pp, low confidence
28Claude Opus 588.6%estimated ± 5.6 pp, low confidence
29GPT-6 Sol88.5%estimated ± 5.6 pp, low confidence
30LongCat-2.588.4%measured
31Kimi K388.3%measured
32GLM-5.388.2%measured
33Claude Mythos 588.0%measured
34DeepSeek V4 Pro87.9%measured
35Hy487.8%estimated ± 5.6 pp, medium confidence
36MiMo-V2.6-Flash87.6%measured
37GPT-5.6 Terra87.4%measured
38Qwen3.8 Max86.6%measured
39GPT-5.586.2%estimated ± 6.7 pp, medium confidence
40Ornith-1.5-397B86.1%measured
41Gemini 3.7 Flash85.8%measured
42Gemini 3.6 Flash85.5%estimated ± 6.7 pp, medium confidence
43Hy4 preview85.4%measured
44Grok 4.685.4%estimated ± 5.6 pp, medium confidence
45Inkling 2 Small85.1%measured
46Step 5 Preview85.0%measured
47GPT-5.6 Luna84.7%measured
48MAI Code 284.6%measured
49Claude Fable 584.3%measured
50GLM-5.3-Flash84.3%measured
51GPT-6 Luna84.2%estimated ± 5.4 pp, low confidence
52Claude Opus 4.783.7%estimated ± 6.7 pp, medium confidence
53Grok 4.583.3%measured
54Ultramarine 1 Flash83.2%measured
55Kimi K2.7 Code82.9%estimated ± 6.7 pp, medium confidence
56Muse Spark 1.282.9%measured
57DeepSeek V4 Flash 073182.7%measured
58Sakana Fugu-Ultra82.1%measured
59Ember-182.0%measured
60SWE-1.781.5%measured
61Ling 3.1 Flash81.2%measured
62GLM-5.281.0%measured
63Ripple-65B Preview80.5%measured
64Claude Sonnet 580.4%measured
65Sakana Fugu80.2%measured
66Beam 1.0 501B-A23B80.1%measured
67Muse Spark 1.180.0%measured
68Particle-179.9%measured
69Beam 1.0 501B-A23B FP878.9%measured
70Qwen3.7 Max78.8%estimated ± 6.7 pp, medium confidence
71MiMo-V2.578.5%estimated ± 6.7 pp, medium confidence
72Beam 1.0 501B-A23B NVFP478.4%measured
73Atria Dawn Preview78.3%measured
74MAI Code 2 Flash78.3%measured
75Ornith-1.0-397B77.5%measured
76Gemini 3.5 Flash76.2%measured
77Composer 2.576.2%estimated ± 6.7 pp, medium confidence
78dots3-note Preview75.1%measured
79Claude Sonnet 4.674.8%estimated ± 6.7 pp, medium confidence
80MiMo-V2.5-Pro74.8%estimated ± 6.7 pp, medium confidence
81Claude Opus 4.874.6%measured
82GLM-5.174.3%estimated ± 6.7 pp, medium confidence
83Jais-3-375B-A23B73.8%measured
84Qwen3.8-27B73.0%measured
85Gemini 3.1 Pro73.0%estimated ± 5.6 pp, medium confidence
86OrcaCyber-Nano-1.072.5%measured
87Seed 2.1 Pro71.0%measured
88Apodex 1.170.8%measured
89GPT-5.4 mini70.6%estimated ± 5.6 pp, medium confidence
90Laguna S 2.170.2%measured
91Gemini 3 Flash69.9%estimated ± 6.7 pp, medium confidence
92Kimi K2.669.4%estimated ± 6.7 pp, medium confidence
93Quasar 438B69.3%measured
94Qwen3.6 Plus68.7%estimated ± 6.7 pp, medium confidence
95Qwen3.7 Plus67.9%estimated ± 6.7 pp, medium confidence
96Ornith-1.5-35B-A3B67.8%measured
97Seed 2.1 Turbo67.6%measured
98Jais-3-MoVA-36B-A4B67.3%measured
99Nemotron 3.5 Super VL 120B A12B66.7%measured
100MiniMax M366.0%measured
101Pokee-Isaac 28B65.1%measured
102Inkling-Small64.7%measured
103Ornith-1.0-35B64.2%measured
104Inkling63.8%measured
105Jais-3-32B63.7%measured
106MAI-Code-1.1-Flash62.9%measured
107Nemotron 3.5 Super VL 120B A12B NVFP460.8%measured
108Step 3.7 Flash59.5%measured
109MiniMax M2.759.2%estimated ± 6.7 pp, medium confidence
110Ling 3.0 Flash57.0%measured
111Solar Pro 457.0%measured
112Nemotron 3 Ultra56.4%measured
113Flint54.9%measured
114Gemini 3.5 Flash-Lite54.0%measured
115Nemotron 3.5 Bolt 74B-A9B BF1653.9%measured
116Ternary Bonsai 2 27B52.8%measured
117Muse Glimmer 30B51.7%measured
118Nemotron 3.5 Bolt 74B-A9B51.0%measured
119Grok 4.2046.8%estimated ± 6.7 pp, low confidence
120Jais-3-7B46.4%measured
121Ornith-1.5-9B46.2%measured
122Claude Haiku 4.545.5%estimated ± 6.7 pp, low confidence
123K-EXAONE 2.043.8%measured
124Ornith-1.0-9B43.1%measured
125Jais-3-3.7B41.2%measured
126Grok 4.339.6%estimated ± 6.7 pp, low confidence
127GPT-5.4 nano38.7%estimated ± 6.7 pp, low confidence
128A.X K236.0%measured
129Mistral Medium 3.5 128B30.5%estimated ± 6.7 pp, low confidence
130Granite 4.2 30B29.2%measured
131Nemotron 3.5 Lightning 30B A3B NVFP423.5%measured
132Granite 4.2 8B20.6%measured
133Mercury 2.517.8%estimated ± 6.7 pp, low confidence
134Mellum2.1-12B-A2.5B-Thinking17.4%measured
135Gemini 3.1 Flash-Lite16.8%estimated ± 6.7 pp, low confidence
136Laguna M.116.8%estimated ± 6.7 pp, low confidence
137LFM3-1.7B12.9%measured
138MiniCPM5-2B8.6%measured
139Laguna XS.23.7%estimated ± 6.7 pp, low confidence
140Command A+0.4%estimated ± 6.7 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General