Benchmarks & research

Jev WebMCP Benchmark Run

A WebMCP benchmark run pairs Jev with Mercury 2.5 and reports solving every task at sharply lower model cost.

Maker-reported (not independently measured by JevMade): 100% of benchmark tasks · about 112× lower model cost than maker's GPT-6 Astra comparison

Source checked 2026-09-25 — opened the primary source directly.