Cognition has cut the price of Devin, its autonomous coding agent, across every mode at once.
The cuts
By the company's own announcement, Devin is now 30 to 40% cheaper in Fusion and Normal modes, 15 to 20% cheaper in Ultra, and up to 70% cheaper in Devin Review, the mode that reviews pull requests. Cognition attributes the reduction to parallel execution and context caching rather than to a change in the underlying models.
The published subscription prices have not changed. What dropped is how much usage each task consumes, so the saving shows up in how far a plan goes, not in the monthly bill itself.
Cheaper and better in the same week
Alongside the price change, Cognition says Devin Fusion now ranks first on FrontierCode 1.1 Extended, at an average cost of $0.60 per task. Fusion is the architecture the company introduced in June: a frontier model plans, a cheaper model executes, and work is routed between them mid-task.
Why this matters beyond Devin
Devin launched as one of the most expensive agent products on the market, and price was the standard objection to it. Removing between a third and two thirds of that cost, without swapping in a weaker model, changes who can justify running one.
The wider pattern is worth noticing. The saving here comes from routing and caching, not from a cheaper model underneath. As the frontier labs cut their own prices in the same week, the agent layer is compounding those cuts with architecture. Costs in this category are falling from two directions at once.
Who should look again
Anyone who priced Devin earlier this year and walked away. The review mode in particular, at up to 70% off, is now in the range where it competes with a human reviewer's time on routine pull requests rather than with another tool. Re-run your own numbers before taking anyone's benchmark on trust, including this one.