Jev vs Claude Opus 5.5: Jev 6-8x faster on computer-use checkout task
On the same checkout task, Jev finished in about 3.5 seconds at roughly 180ms per step versus 21-27 seconds for Opus 5.5, though Claude remains better at tasks needing real reasoning.
Jev (@typesafeai) vs Claude Opus 5.5 (@AnthropicAI) computer use 🥊 Gave both the same checkout task on a local demo store built for testing using the same harness and instructions Jev: done in ~3.5s @ ~180ms a step Opus 5.5: done in 21-27s @ 1.3-2.3s a step Jev consistently 6-8x faster across 4 runs though both finished correctly Claude is still way better at anything that needs real reasoning, but if you know what you're looking for and can pick the next click, a fast model choosing from a list wins