Skip to content
integration bench from HeyMilo
  1. Home
  2. Releases

GPT-6 Astra added

GPT-6 Astra joins the public board at 65.40 Task Score. Same 50 tasks, same scorer, one official attempt per task at xhigh reasoning effort. All 50 attempts are published in full.

10 September 2026

What shipped

  • GPT-6 Astra via Codex at xhigh reasoning effort. 65.40 Task Score, 33 of 50 tasks resolved, third place. Model.
  • Runs. One official attempt per task, 60-minute wall clock, scorer task-score-v4-mandatory-gated. All 50 attempt bundles are published in the same layout as the rest of the dataset.
  • Board. The leaderboard now covers 21 models and 1,050 published attempts. Existing published attempts and scores do not change.

Leaderboard.