Benchmarks / Coding

Terminal-Bench 2.0 (Vals)

State-of-the-art set of difficult terminal-based tasks Scores are the values published by VALS AI. Source benchmark version 2.0.

62models
–Open Index weight
2026-06-04last updated
Vals AIpublisher

Terminal-Bench 2.0 (Vals)

Higher is better · best effort per model

Chart settings
62 of 62 selected

full table below

#ModelEffortScoreCost / taskHarness
1GPT-5.5high73.2%$1.82–
2Opus 4.8max70.0%$0.884–
3Opus 4.7high68.5%$0.976–
4Gemini 3.5 Flashhigh67.4%$1.19–
5Gemini 3.1 Prohigh67.4%$0.516–
6GPT-5.3 Codexhigh64.0%$0.345–
7Muse Spark–59.6%$0.070–
8Sonnet 4.6max59.6%$0.519–
9Qwen 3.7 Maxmax59.2%$0.719–
10Opus 4.5high58.4%$1.18–
11Opus 4.6max58.4%$1.14–
12GPT-5.4high58.4%$0.512–
13Kimi K2.6max57.3%$0.193–
14DeepSeek V4max56.2%$0.180–
15Gemini 3 Prohigh55.1%$0.456–
16GLM 5.1max53.9%$0.158–
17Gemini 3 Flashhigh51.7%$0.169–
18GPT-5.2high51.7%$0.470–
19Qwen 3.6 Maxmax51.7%$0.320–
20GLM 5high49.4%$0.247–
21MiniMax M2.7high47.2%$0.074–
22MiniMax M3max46.1%$0.216–
23Qwen 3.6 Plusmax44.9%$0.146–
24GPT-5.1high44.9%$0.408–
25Qwen 3.6 27Bmax44.9%$0.298–
26GPT-5.4 Minihigh44.9%$0.914–
27Grok 4.3xhigh43.4%$1.12–
28Sonnet 4.5high41.6%$0.751–
29MiniMax M2.5high41.6%$0.085–
30Qwen 3.5 Plusnone41.6%$0.322–
31Grok 4.20xhigh40.4%$0.154–
32Kimi K2.5max40.4%$0.140–
33GPT-5.4 Nanohigh39.9%$0.224–
34Gemma 4 31B IThigh39.3%––
35Haiku 4.5xhigh38.2%$0.328–
36GLM 4.7none38.2%$0.184–
37MiniMax M2.1–37.1%$0.070–
38GPT-5high37.1%$0.477–
39Kimi K2medium37.1%$0.116–
40DeepSeek V3.2high36.0%$0.166–
41Laguna M.1–31.5%––
42Gemini 2.5 Prohigh30.3%$0.472–
43Mistral Medium 3.5high30.3%$4.71–
44Grok 4 Fastnone29.2%$0.034–
45Laguna XS.2–28.1%––
46GLM 4.6none28.1%$0.144–
47Grok 4high28.1%$0.553–
48GPT-5 Minihigh27.0%$0.117–
49Grok 4.1 Fasthigh24.7%$0.056–
50Gemini 3.1 Flash-Litehigh24.7%$0.065–
51Qwen 3.5 Flashmax24.7%$0.110–
52Qwen 3 Maxmax24.7%$0.423–
53DeepSeek V3.1none22.5%$0.502–
54Gemini 2.5 Flashmedium21.3%$0.144–
55gpt-oss-120bhigh19.1%$0.132–
56Trinity Large Thinking–18.0%$0.326–
57Mistral Small 4high16.9%$0.106–
58GPT-4.1high14.6%$1.08–
59Magistral Medium 1.2medium13.5%$0.567–
60Mistral Large 3none9.0%$0.030–
61Command Anone2.2%$2.44–
62Llama 4 Maverickmedium2.2%$0.034–