Vals AI evaluates Mercury Decide: frontier accuracy at lowest cost

Vals AI, which publishes model benchmark results, said on October 8, 2026 in a post on X that it had tested Mercury Decide, a new model from Inception, on the same benchmark it used for TypeSafe AI's Jev.

The stated result is that the model matched frontier accuracy on claim verification, which means judging whether a statement is supported, at the lowest cost Vals has measured so far. Two parts of that sentence are comparisons: which models count as frontier, and what range of costs Vals has measured. The post's text gives no scores, costs or model names, and does not name the benchmark or describe its setup.