Model Routing

Routing multi-agent requests with JEV: 96/97

F Fez @fez_chat · 2026-09-21

EN▶ videoModel RoutingHuman-reviewed entry

Cover is generated by this site (not the creator's media).

WHAT IT DOES

The author calls routing the unglamorous part of a multi-agent workspace: when you ask for help, something has to pick the teammate. Fez uses JEV for that decision.

On their own routing battery the model scored 96/97 with a 184 ms median, at $0.0018 for the whole run. A local Qwen model stays in place as the fallback, so the workspace still answers when JEV is not the right tool.

Reported by the creator on X; the battery is theirs and was not reproduced here.

Summary written by this site

FROM THE ORIGINAL POST

Fez uses @typesafeai's Jev for that. 96/97 on our routing battery, 184 ms median, $0.0018 for the whole run. Local Qwen stays as fallback. — Fez · Original post

JEV ROLE

Routing Choice #routing#agent#accuracy#cost

WHERE IT SITS

HOW IT WORKS

INPUTan incoming request in a multi-agent workspace

JEVRouting

OUTPUTwhich teammate handles it

REPORTED RESULT

Latency184 ms median

Reported cost$0.0018 per run

Reported accuracy96/97 on their routing battery

Reported by the creator, not independently verified.

WHAT THE POST SHOWED

No observable metrics recorded. Metrics are only published when they can be tied to a source link.

SOURCE

↗ Original post

Source checked

Evidence is a status, not a ranking. It says how the claim can be checked — not whether the case is good.