Devin Slashes Usage Costs by Another 30% to 40%, Fusion Tops Its Own Benchmark

Beating AI Express: AI programming firm Cognition has rolled out significant cost reductions for its coding agent Devin. Devin can independently read, modify, test code and complete development tasks. Costs for its Fusion and Normal modes have fallen by 30% to 40%, Ultra mode by 15% to 20%, and Devin Review by up to 70%. Fusion is Devin’s multi-model collaboration mode, where more capable models handle planning and review while lower-cost models take on specific execution tasks. Cognition also optimized its execution framework this time, merging multiple commands into a single call, running independent steps in parallel, and leveraging more context caching to reduce redundant computations and model calls. Alongside the cost cuts, Cognition announced improved self-test results: on its self-built FrontierCode 1.1 Extended benchmark, Devin Fusion scored 68.8 points, with an average cost of $0.60 per task, and currently ranks first.