Devin is now up to 40% more cost-efficient | Devin

Usage costs are reduced by 30-40% in Fusion and Normal modes, 15-20% in Ultra, and up to 70% in Devin Review.
Key points
- Harness updates batch shell commands and execute independent tool calls in parallel to reduce turn counts.
- The harness structure is optimized for prompt caching across multi-turn sessions.
- Devin routes across models including SWE-2, Opus 5.5, and GPT-6 Sol, Astra, and Luna.
- Devin Fusion scored 68.8 on the FrontierCode 1.1 Extended benchmark at an average cost of $0.60 per task.
Why it matters
Developers and teams using Devin can run coding agent tasks and automated pull request reviews with lower token and execution costs.
Sources
- Devin is now up to 40% more cost-efficient | Devin (Devin, Sep 28, 2026, 6:02 PM)