Highlights
- Corrects quota API-equivalent estimates by incorporating finalized subscription-funded Core, Extended, and DigBench usage; estimates are deferred while benchmark accounting is incomplete.
- Fails closed on partial benchmark telemetry and aggregate long-context usage that cannot be priced accurately.
- Separates benchmark setup and model-turn deadlines so slow platform setup no longer consumes model execution time.
Full changelog: v0.11.0...v0.11.1