Terminal-Bench 4.0 (Vals run)

Frontier-difficulty terminal tasks across software, science, ML, operations, hardware and security, run with the Mini-SWE-agent harness. Same benchmark as the official Terminal-Bench 4 board, so it shares its weight. Run by Vals AI.

CategoryAgents
MetricAccuracy (%)
SourceVals AI (evaluator)
LicenseNone stated
Benchmark pagewww.vals.ai
StatusScored
Weight1.1 within Agents
FitDiscrimination 2.85, noise 0.40
LineageShares weight with terminal_bench_4

Results

#ModelProviderResultDate
1 Claude Opus 5.5 Max Anthropic 65.2% 2026-11-09
2 Claude Fable 5.2 High Anthropic 65.2% 2026-11-09
3 Claude Sonnet 5.5 Max Anthropic 64.1% 2026-11-09
4 Kimi K3.1 Pro Max Moonshot AI 63.6% 2026-11-09
5 Gemini 4 Carbon High Google 63.1% 2026-11-09
6 GPT-6 Astra Max OpenAI 59.6% 2026-11-09
7 Claude Fable 5.1 Max Anthropic 58.1% 2026-11-09
8 Muse Wonder Full Max Meta 58.1% 2026-11-09
9 Gemini 4 Argon High Google 57.6% 2026-11-09
10 GPT-6.1 Sol Max OpenAI 55.1% 2026-11-09
11 Claude Opus 5 Max Anthropic 53.5% 2026-11-09
12 GPT-6.1 Terra Max OpenAI 49.5% 2026-11-09
13 Muse Wonder Max Meta 49.5% 2026-11-09
14 GPT-6 Sol Max OpenAI 44.4% 2026-11-09
15 Claude Fable 5 Max Anthropic 41.4% 2026-11-09
16 Hy4 Tencent 39.4% 2026-11-09
17 Muse Ember Max Meta 39.4% 2026-11-09
18 GLM-5.3 Max Zhipu AI 38.9% 2026-11-09
19 GPT-5.6 Sol Max OpenAI 37.9% 2026-11-09
20 GLM-5.3-Flash Max Zhipu AI 37.4% 2026-11-09
21 Claude Haiku 5.5 Max Anthropic 36.4% 2026-11-09
22 Qwen3.8 Max Alibaba 34.3% 2026-11-09
23 Step 5 Preview StepFun 31.8% 2026-11-09
24 DeepSeek V4.1 Flash 1017 Max DeepSeek 31.8% 2026-11-09
25 MiMo-V2.6-Pro Xiaomi 31.3% 2026-11-09
26 Grok 4.7 Extra High xAI 28.8% 2026-11-09
27 Mistral Large 4 High Mistral 27.8% 2026-11-09
28 MiniMax-M3.1-Flash Max MiniMax 25.3% 2026-11-09
29 Muse Spark 1.3 Max Meta 24.7% 2026-11-09
30 MiMo-V2.6-Flash Xiaomi 24.2% 2026-11-09
31 Claude Opus 4.8 Max Anthropic 23.2% 2026-11-09
32 GPT-5.6 Terra Max OpenAI 22.7% 2026-11-09
33 Ember-1 Fireworks 20.7% 2026-11-09
34 Gemini 3.8 Flash High Google 19.7% 2026-11-09
35 DeepSeek V4.1 Flash High DeepSeek 19.7% 2026-11-09
36 DeepSeek V4 Flash 0731 High DeepSeek 18.7% 2026-11-09
37 Grok 4.6 High xAI 17.2% 2026-11-09
38 Kimi K3 Max Moonshot AI 17.2% 2026-11-09
39 DeepSeek V4 Pro 0813 Max DeepSeek 14.1% 2026-11-09
40 GPT-6 Luna Max OpenAI 13.6% 2026-11-09
41 Inkling 2 Small Thinking Machines 12.6% 2026-11-09
42 Gemini 3.7 Flash High Google 12.1% 2026-11-09
43 GPT-5.6 Luna Max OpenAI 11.6% 2026-11-09
44 DeepSeek V4 Pro Max DeepSeek 11.1% 2026-11-09
45 Claude Sonnet 5 Max Anthropic 10.6% 2026-11-09
46 Muse Spark 1.3 Extra High Meta 10.6% 2026-11-09
47 Grok 4.5 High xAI 8.6% 2026-11-09
48 Hy4 preview Tencent 8.1% 2026-11-09
49 Muse Spark Open Extra High Meta 6.6% 2026-11-09
50 Gemini 3.5 Flash High Google 6.1% 2026-11-09
51 Muse Spark 1.2 Extra High Meta 6.1% 2026-11-09
52 Beam 1.0 Reflection 4.0% 2026-11-09
53 Mercury 3 High Inception 3.0% 2026-11-09
54 GPT-5.4 mini Extra High OpenAI 2.5% 2026-11-09
55 Gemini 3.1 Pro Preview High Google 2.5% 2026-11-09
56 MiniMax-M3 MiniMax 1.0% 2026-11-09
57 Inkling Thinking Machines 0.5% 2026-11-09
58 MiMo-V2.5-Pro Xiaomi 0.5% 2026-11-09
59 Inkling Small Thinking Machines 0.0% 2026-11-09
60 Nemotron 3.5 Super VL 120B A12B High NVIDIA 0.0% 2026-11-09
61 Mercury 2.5 High Inception 0.0% 2026-11-09