DoorDash Releases DashBench to Evaluate Multi-Model AI Code Review
Commentary on X
Highest ranked
Because of DashBench, we’re able to quickly discern which model combinations yield the best results and at the best cost. For example, with DashBench we’ve seen Kimi K2.6 + Fable 5 vastly outperform our current Sonnet 4.6 + Opus 4.8 harness at a cheaper cost. Having our own benchmark allows us to build confidence in leveraging the frontier intelligence and open-weight models without compromising on enterprise outcomes.
@AIatDoorDash Going from 30% to 53.6% recall with a multi-model setup is a meaningful jump, and shipping DashBench as a public benchmark is the right move for the whole space.
@AIatDoorDash Going from 30% to 53.6% recall just by changing the review architecture is a bigger deal than most people will notice.
@AIatDoorDash You can fix all the errors in your codebase but how does that solve the human error introduced by black scammers and thieves?
