Factory has released Droid 1.0, a new model-independent coding agent harness designed to utilize programmatic tool calling by default. According to reporting by MarkTechPost, the system runs on script mode. During internal tests using Terminal-Bench 4, Droid achieved a score of 65.2 percent when paired with Claude Opus 5.5, with an average cost of $4.73 for each successfully passed task.
The Shift Toward Programmatic Tool Calling
Coding agent harnesses like Droid 1.0 represent an ongoing effort in software development to make AI models more autonomous by structuring how they interact with developer tools and terminals. By keeping the harness model-independent, developers can theoretically swap out underlying large language models depending on performance and cost requirements. Script mode acts as the foundational environment for executing these programmatic calls efficiently.
Open Questions and Future Metrics
Despite the initial benchmark figures shared in the release, several details remain unknown. The announcement did not specify how Droid 1.0 performs with models other than Claude Opus 5.5, nor did it provide broader test results across alternative coding benchmarks. Furthermore, hardware requirements and potential latency issues were not addressed in the available information, leaving developers waiting for independent evaluations.
As automated software engineering tools continue to evolve, third-party testing will help determine whether Droid 1.0 maintains its cost and performance efficiency in diverse production environments.
Lead image: Rafael Minguet Delgado / Pexels
