02 — Run · Round 2 · Round robin · r2-rr-sonnet5-gpt56terra-g1
Claude Sonnet 5 vs GPT-5.6 Terra
GPT-5.6 Terra destroyed the enemy base after 6.1 game-minutes · 8.4 min wall clock
LOSS russia · west
Claude Sonnet 5
anthropic/claude-sonnet-5
Value destroyed$3,600Value lost$13,300
Units killed / lost20 / 4Buildings killed / lost0 / 13
Peak army$700Orders issued73
Decision turns (failed)47 (0)Mean latency4.7sModel cost$0.285
WIN germany · east
GPT-5.6 Terra
openai/gpt-5.6-terra
Value destroyed$13,300Value lost$3,600
Units killed / lost4 / 20Buildings killed / lost13 / 0
Peak army$3,100Orders issued50
Decision turns (failed)47 (0)Mean latency3.4sModel cost$0.329
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 | gpt56terra won: base destroyed |
| pass | code_check | decisions-logged | turns per slot: {'Multi0': 47, 'Multi1': 47} |
| pass | code_check | llm-turns-valid | failed/total per slot: {'Multi0': '0/47', 'Multi1': '0/47'} |
| pass | code_check | video-decodes | duration 63.4s |
| pass | code_check | video-complete | video ends at tick 9201 of 9201 |
| pass | code_check | limits-respected | ended after 8.4 wall-minutes (budget 30.0 min): base destroyed |
| pass | checklist | notes-explain-win | Notes section cites last thoughts from both models and explains how GPT-5.6 Terra won via economic dominance and base destruction |
| pass | checklist | players-match-request | result.json shows Multi0=sonnet5 (anthropic/claude-sonnet-5) and Multi1=gpt56terra (openai/gpt-5.6-terra) matching the requested players |
| pass | checklist | no-silent-outage | 0 LLM failures out of 47 calls per player, regular 200-tick intervals with no gaps, no errors 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.