Factory's coding harness reports 65.2% at $4.73 per task

single source· 1 articles · confidence: low · first seen 2026-10-31 10:38 UTC

What this means for you

Nothing to act on unless you are already shopping for a coding harness. The 65.2% comes from the vendor's own runs, with no evaluation date and no rival harness on the same tasks; the $4.73 is charged per passing task rather than per call, which is the figure worth asking about.

Factory has released Droid 1.0, a coding agent harness (the scaffolding that hands a model a task and runs the tools it asks for) that works with more than one model and makes programmatic tool calling the default, so the model issues calls as code rather than one at a time. On Factory's own Terminal-Bench 4 runs it scored 65.2% with Claude Opus 5.5, at $4.73 per passed task. The article gives no evaluation date, no second harness measured on the same tasks, and no price or availability details for Droid itself.

Models in this story

Key facts

  • ·Factory released Droid 1.0, a coding agent harness that is not tied to a single model. source
  • ·Programmatic tool calling is the default mode in Droid 1.0. source
  • ·Droid scored 65.2% on Factory's own Terminal-Bench 4 runs using Claude Opus 5.5. source
  • ·Those runs cost $4.73 per passed task. source
  • ·No evaluation date is given for the Terminal-Bench 4 score. source

What the sources say

  • MarkTechPost — Short release write-up carrying the vendor's benchmark score and per-task cost, with no outside comparison.

Sources

The original reporting. Follow these — they did the work.

← the wire