Clinical benchmarks show Astra is incremental, still trailing Anthropic's frontier

danielmckinn0n · x · 2026-09-06

The author tested OpenAI's Astra on RareBench, a clinical genetics benchmark, and concluded it's an incremental improvement over Sol rather than a new paradigm, still behind Anthropic's Fable 5.1.

Key points:

Original post →

More from Models

Models channel →