| Paper | Code | Accuracy | ModelName | ReleaseDate |
|---|---|---|---|---|
| Introducing Claude Fable 5.1 and Claude Mythos 5.1 | 69.0 | Claude Fable 5.1 (max effort, with fallbacks) | 2026-09-01 | |
| Introducing Claude Opus 5.5 | ✓ Link | 67.7 | Claude Opus 5.5 (max effort) | 2026-09-22 |
| Introducing Claude Sonnet 5.5 | ✓ Link | 65.6 | Claude Sonnet 5.5 (max effort) | 2026-09-28 |
| Introducing Claude Haiku 5.5 | 60.3 | Claude Haiku 5.5 (max effort) | 2026-10-07 | |
| WorkRover: Scaling Long-Horizon Cowork Agent Training Via Recursive Environment Evolution | 48.87 | WorkRover-35B-A3B (Claude Code) | 2026-10-13 |