Compare seven models on your own Decisions jobs
Your last 30 days of Decisions asks, priced on Laya, Jev, DeepSeek, Haiku, Sol, Sonnet and Opus, beside our screen test and each lab's Terminal-Bench score.
The comparison in Decisions now takes every ask your agents made in the last 30 days, job by job, and prices it at each model's list price. Pick a job to see what it alone would cost. The model in use is marked, and the line under the table follows it when you change the Decisions model.
Two quality columns sit beside the prices: our own 104-screen test, which only the Decisions models ran, and Terminal-Bench 4.0 as each lab publishes it. Where a lab publishes none, the cell says so.
The TypeSafe card now counts the asks sent to Jev, and compaction and context ranking asks now count toward Decisions usage.