evals
Run evals on Jetty

02 — Run · Round 1 · Round robin · rr-sonnet5-gpt56terra-g3

Claude Sonnet 5 vs GPT-5.6 Terra

GPT-5.6 Terra destroyed the enemy base after 5.3 game-minutes · 8.2 min wall clock

LOSS russia · west
Claude Sonnet 5
anthropic/claude-sonnet-5
Value destroyed$200Value lost$11,850 Units killed / lost2 / 3Buildings killed / lost0 / 12 Peak army$300Orders issued45 Decision turns (failed)40 (0)Mean latency4.7sModel cost$0.242
WIN germany · east
GPT-5.6 Terra
openai/gpt-5.6-terra
Value destroyed$11,850Value lost$200 Units killed / lost3 / 2Buildings killed / lost12 / 0 Peak army$2,700Orders issued42 Decision turns (failed)40 (0)Mean latency3.3sModel cost$0.287

Evaluation

passchecks 7/7 verdict computed from the gating checks

Round-1 run: written before the runbook moved to evals v2, so these are the older self-reported checks.

StatusKindCheckDetail
passcode_checkresult-jsonresult json
passcode_checkhas-winnerhas winner
passcode_checkdecisions-loggeddecisions logged
passcode_checkreplay-savedreplay saved
passcode_checkvideo-renderedvideo rendered
passcode_checkno-llm-outageno llm outage
passcode_checkvideo-completevideo complete

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?

Get Started Book a 15-minute walkthrough
Connect your agent to Jetty →