Cognition Updates Devin Fusion on FrontierCode 1.1
Cognition shared benchmark results for its AI coding agent.
Cognition posted on X that its Devin Fusion AI coding agent now scores 4% higher while costing 27% less per rollout on FrontierCode 1.1. The company attributed the gains to improvements in both the harness and models. A link to the announcement includes a photo. Other visible replies mention orchestration of frontier and smaller models for cost savings. The post promotes trying Devin Fusion in Devin Cloud.
Thanks to improvements in both the harness and models, Devin Fusion is now 4% more intelligent and 27% less expensive on FrontierCode 1.1.
Combined views
65.2K
Reactions from ranked influencers
3 postsTry Devin Fusion today in Devin Cloud: http://app.devin.ai/signup
Last week, OpenAI massively cut the cost of Luna models. Devin Fusion orchestrates frontier models alongside smaller execution models like Luna This means even users paying for frontier performance get cost savings when the price of sub-frontier intelligence goes down
Thanks to improvements in both the harness and models, Devin Fusion is now 4% more intelligent and 27% less expensive on FrontierCode 1.1.
Our research team is hill climbing on the pareto frontier of intelligence vs. cost efficiency with Fusion. With tokenmaxxing era now over, hybrid architectures and routers are becoming the default for many large organizations.
Thanks to improvements in both the harness and models, Devin Fusion is now 4% more intelligent and 27% less expensive on FrontierCode 1.1.
