TeamBench - Restricted: leaderboard
Metric: Pass rate (%) on the 90-task stratified TeamBench leaderboard subset (operating-system-enforced role separation in Docker sandboxes; a run passes when every check of the task's deterministic shell-script grader passes, and a run failing only the attestation check counts as a pass), temperature 0, with one agent holding only the Executor's capabilities (it edits the workspace and runs commands but cannot read the full specification or write the attestation); higher is better. Source: arxiv.org. Saturation forecast: Around 2031. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.4 | 35.6 |
| 2 | Claude Opus 4.7 | 33.3 |
| 3 | Claude Haiku 4.5 | 31.1 |
| 4 | Claude Sonnet 4.6 | 27.8 |
| 5 | Gemma 4 31B | 25.6 |
| 6 | GPT-5.4 Mini | 23.3 |
| 7 | Gemini 3.1 Pro (Preview) | 22.2 |
| 8 | Gemini 3.1 Flash Lite | 21.1 |
| 9 | Gemini 3 Flash | 18.9 |
| 10 | GPT-OSS-20B | 17.8 |
Interactive version: theaggregate.ai/benchmark?slug=teambench-restricted · How It Works · Data refreshed daily, snapshot 2026-10-07.