Computer Anthology Terminal Tasks (Codex CLI): leaderboard
Vetto's Computer Anthology Terminal Tasks v1.0 run under the Codex CLI agent scaffold — 100 held-out terminal jobs graded by per-task sandboxed verifiers; pass@1 over 5 trials per task. Reported per model+scaffold pair because the scaffold moves the score substantially.
Metric: pass@1 (%). Source: vetto.ai. Saturation forecast: Around October 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Astra (High) | 65.6 | #6 (GPT-6 Astra) |
| 2 | GPT-5.6 Sol (High) | 58 | #22 (GPT-5.6 Sol) |
| 3 | GPT-5.5 (High) | 51.6 | #33 (GPT-5.5) |
| 4 | GPT-5.4 (High) | 38 | #83 (GPT-5.4) |
| 5 | GPT-5.4 Mini (High) | 21.8 | #228 (GPT-5.4 Mini) |
No result here: #3 Claude Opus 5.5, #7 GPT-6.1 Sol, #9 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=computer-anthology-terminal-tasks-codex-cli · How It Works · Data refreshed daily, snapshot 2026-11-05.