Benchmark

Terminal-Bench 2.0

Measures: 89 tasks an agent completes in a Docker terminal (building, configuring, debugging, data work), each checked by a test script; version 2.0 of Terminal-Bench, whose scores are not comparable with 4.0.

Maintainer
Laude Institute and Stanford (Terminal-Bench team)
Introduced
2025
Status
Superseded
Saturation
Near-ceiling
Typical provenance
Mixed
In scoring set
No

PaperOfficial leaderboard

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

Leaderboard
ModelAccuracy (mean over trials)ProvenanceSourceEvaluated
GPT-5.5
OpenAI · Unknown effort
84.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.5
OpenAI · Unknown effort
83.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.5
OpenAI · Unknown effort
82.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.5
OpenAI · Unknown effort
82.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.4
OpenAI · Unknown effort
81.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.7
Anthropic · Unknown effort
80.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3.1 Pro80.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
79.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex78.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3.1 Pro78.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 3.1 Pro78.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex77.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
76.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex75.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex75.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
75.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex75.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3.1 Pro74.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
74.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex74.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
72.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
71.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex71.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex70.3%Source not statedEpoch AI via Epoch AI—
Claude Opus 4.669.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
69.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro69.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.3-Codex69.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
66.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
66.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2-Codex66.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.5
OpenAI · Unknown effort
66.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro65.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Unknown effort
64.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Medium effort
64.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5.3-Codex64.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Flash64.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
63.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.563.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
62.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Unknown effort
62.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Medium effort
62.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 3 Pro62.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro62.2%Source not statedEpoch AI via Epoch AI—
Gemini 3 Pro61.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex-Mini61.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro61.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Unknown effort
60.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex-Max60.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3.1 Pro59.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
59.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · 128K effort
59.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
58.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.6
Anthropic · Unknown effort
58.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex57.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
57.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex57.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok 4.2057.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok 4.2057.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro56.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Pro56.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
54.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.2
OpenAI · Unknown effort
54.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex53.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.6
Anthropic · Unknown effort
53.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.653.4%Source not statedEpoch AI via Epoch AI—
GLM-552.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
52.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
51.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.5
Anthropic · Unknown effort
51.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Flash51.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Flash51.0%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5
OpenAI · Unknown effort
49.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5
OpenAI · Medium effort
49.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5.1
OpenAI · Medium effort
47.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5.1
OpenAI · Medium effort
47.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5.1
OpenAI · Unknown effort
47.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 3 Flash47.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
46.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
MiniMax-M2.745.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5-Codex44.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5-Codex44.3%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5
OpenAI · Unknown effort
43.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5
OpenAI · Medium effort
43.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5-Codex43.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5-Codex43.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Kimi K2.543.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
43.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex-Mini43.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex-Mini43.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
MiniMax-M2.742.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
42.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.542.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
42.7%Source not statedEpoch AI via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
42.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
MiniMax-M2.542.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.542.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
42.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
42.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Sonnet 4.542.5%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
MiniMax-M2.542.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5-Codex41.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5-Codex41.3%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Sonnet 4.540.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Sonnet 4.5
Anthropic · Unknown effort
40.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
DeepSeek-V3.239.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.138.0%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.1
Anthropic · Unknown effort
38.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.136.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5.1-Codex36.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.1
Anthropic · Unknown effort
36.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5.1-Codex36.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
MiniMax-M2.136.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Kimi K2 Thinking35.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Kimi K2 Thinking35.7%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Haiku 4.5
Anthropic · Unknown effort
35.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5
OpenAI · Medium effort
35.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5
OpenAI · Unknown effort
35.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.135.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Opus 4.1
Anthropic · Unknown effort
35.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.1
Anthropic · Unknown effort
34.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
34.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Opus 4.134.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
34.8%Source not statedEpoch AI via Epoch AI—
GPT-5
OpenAI · Unknown effort
33.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5
OpenAI · Medium effort
33.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GLM-4.733.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GLM-4.733.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Pro32.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Pro32.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 mini
OpenAI · Medium effort
31.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
31.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
MiniMax-M230.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.5
Anthropic · Unknown effort
29.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.529.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
29.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
MiniMax-M2.129.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 mini
OpenAI · Medium effort
29.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Haiku 4.5
Anthropic · Unknown effort
28.3%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.528.3%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Claude Haiku 4.5
Anthropic · Unknown effort
27.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.527.5%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Grok 427.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Grok 427.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Qwen3-Coder-480B-A35B-Instruct27.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Pro26.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Pro26.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok Code Fast 125.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok Code Fast 125.8%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Grok 425.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Qwen3-Coder-480B-A35B-Instruct25.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Grok 425.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Qwen3-Coder-480B-A35B-Instruct25.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GLM-4.624.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GLM-4.624.5%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
24.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 mini
OpenAI · Medium effort
24.0%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Qwen3-Coder-480B-A35B-Instruct23.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Qwen3-Coder-480B-A35B-Instruct23.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok 423.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok 423.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Qwen3.6-35B-A3B23.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 mini
OpenAI · Unknown effort
22.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 mini
OpenAI · Medium effort
22.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Unknown effort
21.8%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Pro19.6%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Pro19.6%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
gpt-oss-120b18.7%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
gpt-oss-120b
OpenAI · Unknown effort
18.7%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Flash17.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Flash17.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Flash16.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Flash16.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Flash16.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Pro16.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Flash16.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Pro16.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Gemini 2.5 Flash15.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Gemini 2.5 Flash15.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
gpt-oss-120b14.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
Grok Code Fast 114.2%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
gpt-oss-120b
OpenAI · Unknown effort
14.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Grok Code Fast 114.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.5
Anthropic · Unknown effort
13.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Claude Haiku 4.513.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Medium effort
11.5%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Unknown effort
11.5%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 nano
OpenAI · Medium effort
9.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Unknown effort
9.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
Qwen3.5-9B9.2%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 nano
OpenAI · Medium effort
7.9%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Unknown effort
7.9%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
GPT-5 nano
OpenAI · Medium effort
7.0%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
GPT-5 nano
OpenAI · Unknown effort
7.0%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
gpt-oss-20b3.4%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
gpt-oss-20b
OpenAI · Unknown effort
3.4%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
gpt-oss-20b3.1%Third partyTerminal-Bench v2 Leaderboard via Epoch AI—
gpt-oss-20b
OpenAI · Unknown effort
3.1%Third partyhttps://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI—
  • Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.

Sources