Vibeleaderboard
← Benchmark guide

Model results

Terminal-Bench

Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.

Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.

Reviewed recommendation and limits

Start with Terminal-Bench 4.0 for terminal-based agent work. Compare the same task release, agent and resource settings.

Terminal-Bench 4.0.0 on Harbor. Preserve agent, model, reasoning effort, task resources, trial count and resolution-rate uncertainty. Earlier 2.0, 2.1 and 3.0 results remain separate.. Imported groups below may use different protocols; this review does not certify each run. Public solutions, reward hacking and task-specific tuning can inflate results. Earlier editions and vendor-reported runs do not establish the current reviewed protocol. Terminal completion is different from repository issue repair.

Publisher results Terminal-Bench 4.0 Leaderboard integrity policy Contributors
← All evaluation groups

Terminal-Bench 2.0

Source
Terminal-Bench via Epoch AI
Version
2.0
Measurement date
Not reported
Metric
% · Higher is better
Evaluation scope
Agent Terminus 2 · result date not supplied

Positions are withheld because at least one result lacks explicit independent provenance or a documented evaluation scope. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

  1. GPT-5.3-Codex

    openai/gpt-5.3-codex-20260224

    64.7%

    View modelEvidence aggregated

  2. GLM 5

    z-ai/glm-5-20260211

    52.4%

    View modelEvidence aggregated

  3. Gemini 3 Flash Preview

    google/gemini-3-flash-preview-20251217

    51.7%

    View modelEvidence aggregated

  4. GPT-5 Codex

    openai/gpt-5-codex

    43.4%

    View modelEvidence aggregated

  5. Kimi K2.5

    moonshotai/kimi-k2.5-0127

    43.2%

    View modelEvidence aggregated

  6. Claude Sonnet 4.5

    anthropic/claude-sonnet-4-5

    42.8%

    View modelEvidence aggregated

  7. MiniMax M2.5

    minimax/minimax-m2.5-20260211

    42.2%

    View modelEvidence aggregated

  8. Claude Opus 4.1

    anthropic/claude-opus-4-1

    38%

    View modelEvidence aggregated

  9. GPT-5.1-Codex

    openai/gpt-5.1-codex-20251113

    36.9%

    View modelEvidence aggregated

  10. Kimi K2 Thinking

    moonshotai/kimi-k2-thinking-20251106

    35.7%

    View modelEvidence aggregated

  11. GLM 4.7

    z-ai/glm-4.7-20251222

    33.4%

    View modelEvidence aggregated

  12. Gemini 2.5 Pro

    google/gemini-2.5-pro

    32.6%

    View modelEvidence aggregated

  13. MiniMax M2

    minimax/minimax-m2

    30%

    View modelEvidence aggregated

  14. MiniMax M2.1

    minimax/minimax-m2.1

    29.2%

    View modelEvidence aggregated

  15. Claude Haiku 4.5

    anthropic/claude-haiku-4-5

    28.3%

    View modelEvidence aggregated

  16. Kimi K2 0711

    moonshotai/kimi-k2

    27.8%

    View modelEvidence aggregated

  17. GLM 4.6

    z-ai/glm-4.6

    24.5%

    View modelEvidence aggregated

  18. Qwen3 Coder 480B A35B

    qwen/qwen3-coder-480b-a35b-07-25

    23.9%

    View modelEvidence aggregated

  19. gpt-oss-120b

    openai/gpt-oss-120b

    18.7%

    View modelEvidence aggregated

  20. Gemini 2.5 Flash

    google/gemini-2.5-flash

    16.9%

    View modelEvidence aggregated

  21. gpt-oss-20b

    openai/gpt-oss-20b

    3.1%

    View modelEvidence aggregated
Terminal-Bench model results | VibeLeaderboard