Vals AI

Terminal-Bench 1.0

State-of-the-art set of difficult terminal-based tasks

In index
Agentic
Source ↗

Models45

Top result63.8%GPT-5.2

Top-3 spread3.8 pts

Updated12 Jan 2026

Result and cost

Results

46 results

#ModelResultCostTime

01

GPT-5.263.8%$1.424 min

02

Claude Sonnet 4.5 Thinking61.3%$0.6619 min

03

Gemini 3 Flash Preview60.0%$0.1262 min

04

GPT-5 Codex58.8%$1.3811 min

05

Claude Opus 4.5 Thinking57.5%$0.9615 min

05

GPT-5.1-Codex57.5%$0.12116 min

07

Claude Opus 4.556.3%$0.7443 min

08

GPT-5.1-Codex-Max53.8%$1.0212 min

09

Gemini 3 Pro51.3%$0.3188 min

10

GLM-4.750.0%$0.2239 min

10

DeepSeek V3.250.0%$0.0257 min

10

Claude Haiku 4.5 Thinking50.0%$0.266 min

13

GPT-548.8%$0.10116 min

14

GPT-5.147.5%$0.10211 min

15

Claude Sonnet 4 Thinking45.0%$2.0212 min

16

Devstral43.8%$0.4258 min

17

GLM-4.642.5%$0.2862 min

18

MiniMax M2.141.3%$0.0637 min

18

Gemini 2.5 Pro41.3%$0.6646 min

18

GLM-4.541.3%$0.1973 min

18

DeepSeek V3.141.3%$0.39111 min

22

DeepSeek V3.2 (Thinking)40.0%$0.0486 min

22MAKimi K2 Thinking40.0%$0.3352 min

22

Labs Devstral Small40.0%$0.1355 min

25

Grok 438.8%$7.3115 min

25MAKimi K2 Instruct38.8%$0.3352 min

27

Qwen3 Max36.3%$0.4815 min

27

Qwen3 Max36.3%$0.7153 min

29

GPT-4.133.8%$1.295 min

30

Gemini 2.5 Flash (Preview)31.3%$0.13479 s

31

GPT-5 mini30.0%$0.36625 min

32

Grok 4.1 Fast (Reasoning)28.8%<$0.014 min

32

Magistral Medium28.8%$1.2312 min

34

Grok 4 Fast (Reasoning)27.5%$0.1882 min

35

Gemini 2.5 Flash Preview Thinking26.3%$0.1332 min

36

GPT-OSS 120B22.5%$0.03954 s

37

Grok 4.1 Fast Non Reasoning21.3%$0.0272 min

37

Mistral Large21.3%$0.0573 min

37

Gemini 2.5 Flash Thinking21.3%$0.4484 min

40

Grok Code Fast 120.0%$0.23885 s

41

Grok 4 Fast Non Reasoning18.8%$0.31165 s

42

Llama4 Maverick Instruct Basic17.5%$0.03555 s

43

Magistral Small15.0%$0.22110 min

44

DeepSeek-R113.8%$0.362 min

45COCommand A6.3%$2.55 min

45ALJamba Large 1.76.3%$4.0210 min