02 — Run · Round 2 · Round robin · r2-rr-jev-gpt6luna-g2
GPT-6 Luna vs Jev Router
Jev Router destroyed the enemy base after 11.5 game-minutes · 15.6 min wall clock
LOSS ukraine · west
GPT-6 Luna
openai/gpt-6-luna
Value destroyed$9,750Value lost$24,700
Units killed / lost35 / 29Buildings killed / lost2 / 9
Peak army$5,750Orders issued46
Decision turns (failed)87 (0)Mean latency3.3sModel cost$0.032
WIN russia · east
Jev Router
typesafe/jev-router
Value destroyed$24,700Value lost$9,750
Units killed / lost29 / 35Buildings killed / lost9 / 2
Peak army$5,900Orders issued156
Decision turns (failed)87 (15)Mean latency5.6sModel cost$0.173
Evaluation
passchecks 10/10 verdict computed from the gating checks
| Status | Kind | Check | Detail |
|---|---|---|---|
| pass | step | environment | Environment check (snapshot, engine, content, key) |
| pass | step | launch | Launch run-match as a detached process |
| pass | step | supervise | Supervise in the foreground until run-match exits |
| pass | step | render | Render the 1080p recording |
| pass | step | notes | Append Notes to summary.md |
| pass | code_check | outputs-exist | all present |
| pass | code_check | game-has-winner | jev won: base destroyed |
| pass | code_check | decisions-logged | turns per slot: {'Multi0': 87, 'Multi1': 87} |
| pass | code_check | llm-turns-valid | failed/total per slot: {'Multi0': '0/87', 'Multi1': '8/87'} |
| pass | code_check | video-decodes | duration 117.3s |
| pass | code_check | video-complete | video ends at tick 17296 of 17296 |
| pass | code_check | limits-respected | ended after 15.6 wall-minutes (budget 30.0 min): base destroyed |
| pass | checklist | notes-explain-win | summary.md Notes section cites final thoughts from both GPT-6 Luna and Jev Router explaining the decisive army-size and economic disparity |
| pass | checklist | players-match-request | result.json records Multi0=gpt6luna (openai/gpt-6-luna) and Multi1=jev (typesafe/jev-router) matching the requested west=gpt6luna east=jev |
| pass | checklist | no-silent-outage | Jev Router had 8 LLM failures out of 87 turns (9.2%); GPT-6 Luna had 0 failures; no long silent stretches in run.log |
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.