Factory released Droid 1.0 — software shell for coding, not bound to a single model, and using tool invocations as a standard mode. This is reported by MarkTechPost.
In Factory's own tests on Terminal-Bench 4, Droid achieved 65,2% with Claude Opus 5.5; the cost was 4,73 dollars per one successfully completed task. The available material does not include methodological details or independent verification of these metrics.
editorial commentary
Why it matters
Likely consequence — increased competition around tools that allow changing the model used without a full reshaping of the workflow. The next observable signal will be independent results on Terminal-Bench 4 and data on supported models and real task costs. Substantial uncertainty remains due to lack of primary material and full test descriptions.