Factory Launches Droid 1.0, Tops Terminal-Bench With 57.6%/65.2% Pass Rates
madhavsinghal_ · x · 2026-10-30
Factory AI introduced Droid 1.0, the new engine behind its coding agent, claiming the top spot on Terminal-Bench.
Benchmarks shown per underlying model (pass rate vs. cost per passed task):
- GPT-6.1 SoL: Droid leads at $0.79 with 57.6%, beating Codex (54.0%), OpenCode (51.0%), and Devin (45.0%); Cursor excluded since it has no access to OpenAI models.
- Opus 5.5: Droid hits 65.2% at $4.73 — highest quality and lowest cost — ahead of OpenCode and Cursor (62.1% each), Devin (59.6%), and Claude Code (57.6%).
Positioned as both more accurate and cheaper than rivals for complex projects and everyday development.
Related event: Factory Launches Droid 1.0, Topping Terminal-Bench at Lowest Cost(3 posts)→
More from coding & agent
- The Infinite Work Machine: buyer bots vs seller bots turn AI efficiency gains into more work — Ididit-forthecookie · 2026-10-30
- Seroter's reading list: agent identity management, AI code review fails, skills as products — rseroter · 2026-10-30
- Supabase shares Multigres build story and opens early access signups — dshukertjr · 2026-10-30
- NVIDIA shows developers building Omniverse simulations with frontier AI agents — RevLebaredian · 2026-10-30
- Dev builds MCP server to visualize and manage parallel coding agents in a web UI — Shreevenkr · 2026-10-30
- Claude Projects is Anthropic's best idea and worst implementation, argues power user — Many_Region8176 · 2026-10-30