evals
Create your eval

02 — Run · Round 2 · Round robin · r2-rr-jev-gpt6luna-g3

Jev Router vs GPT-6 Luna

Jev Router destroyed the enemy base after 9.3 game-minutes · 16.6 min wall clock

WIN england · west
Jev Router
typesafe/jev-router
Value destroyed$19,000Value lost$4,700 Units killed / lost24 / 17Buildings killed / lost9 / 0 Peak army$7,550Orders issued94 Decision turns (failed)71 (20)Mean latency6.0sModel cost$0.118
LOSS germany · east
GPT-6 Luna
openai/gpt-6-luna
Value destroyed$4,600Value lost$18,900 Units killed / lost16 / 23Buildings killed / lost0 / 9 Peak army$2,300Orders issued48 Decision turns (failed)71 (0)Mean latency3.0sModel cost$0.024

Evaluation

passchecks 10/10 verdict computed from the gating checks

StatusKindCheckDetail
passstepenvironmentEnvironment check (snapshot, engine, content, key)
passsteplaunchLaunch run-match as a detached process
passstepsuperviseSupervise in the foreground until run-match exits
passsteprenderRender the 1080p recording
passstepnotesAppend Notes to summary.md
passcode_checkoutputs-existall present
passcode_checkgame-has-winnerjev won: base destroyed
passcode_checkdecisions-loggedturns per slot: {'Multi0': 71, 'Multi1': 71}
passcode_checkllm-turns-validfailed/total per slot: {'Multi0': '13/71', 'Multi1': '0/71'}
passcode_checkvideo-decodesduration 95.4s
passcode_checkvideo-completevideo ends at tick 14001 of 14001
passcode_checklimits-respectedended after 16.6 wall-minutes (budget 30.0 min): base destroyed
passchecklistnotes-explain-winNotes section cites final thoughts from both Jev Router and GPT-6 Luna and explains how Jev Router won by economy and aggression
passchecklistplayers-match-requestresult.json confirms Multi0=jev (Jev Router/typesafe/jev-router) and Multi1=gpt6luna (GPT-6 Luna/openai/gpt-6-luna) matching the requested players
passchecklistno-silent-outage13 isolated LLM failures for Multi0 (18.3% of 71 calls, under 20% threshold) visible in run.log as delayed turns; no consecutive empty stretch; Multi1 had 0 failures

Decision timeline

What each model saw as its situation and what it ordered, every 8 game-seconds. Click a turn to jump the video there.

Ready to stop guessing if your outputs are good?

Create your eval Book a 15-minute walkthrough
Connect your agent to Jetty →