BrowseComp - Sonnet5 multi-agent summary condition
Search
Sonnet5 system-card summary reports a multi-agent BrowseComp score without specifying that run's agent count or scaffold. Diagnostic until the implementation is established; not substituted for the documented single-agent10M-token evaluation.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 5 | 86.6 |