Devin adds Haiku 5.5, claiming Sonnet 5 results at an eighth of the cost
- Source
- x.com
- Date
Claude Haiku 5.5 is now available in Devin. On FrontierCode 1.1, Haiku 5.5 scores 58.4%. That puts it ahead of Sonnet 5 at about an eighth of the cost per task. It is also an excellent sidekick model in Fusion. Try it today with Opus 5.5 as the Lead in Devin Desktop and CLI.

Context
Cognition, maker of the Devin , says Claude Haiku 5.5 beats Sonnet 5 on its coding while costing about an eighth as much per task. In an X post announcing that Haiku 5.5 from Anthropic is now available in Devin, Cognition reports a 58.4% score on FrontierCode 1.1 and puts that ahead of Sonnet 5. If that holds, a team running many agent tasks could get similar or better results from a much cheaper model.
The post gives no Sonnet 5 score and does not explain how cost per task was worked out. In an earlier post about Sonnet 5.5, Cognition reported Sonnet 5 at 56.2% on FrontierCode 1.1 Main. That would make the lead about two points if both posts used the same variant, but this post does not name one. Both the score and the cost ratio are Cognition's own measurements. Cognition also recommends Haiku 5.5 as the sidekick model in Fusion, a Devin where it works alongside Opus 5.5 as the Lead, in Devin Desktop and the . The post does not describe how the two models split the work.
- Agent mode — A mode in which one agent can inspect, edit, run commands, and iterate toward an outcome.
- Benchmark — A fixed set of tasks with a stated way of scoring them, used to compare models or setups.
- Setup — The combination of a model, a harness, its connections and the path the task's data takes, considered together for a task.
- CLI tool — A command-line capability an agent can discover and compose through shell commands.
Checking sign-in…
Loading comments…






