evals
Run evals on Jetty

02 — Run · Round 2 · Round robin · r2-rr-sonnet5-gpt56terra-g3

Claude Sonnet 5 vs GPT-5.6 Terra

GPT-5.6 Terra destroyed the enemy base after 4.8 game-minutes · 7.3 min wall clock

LOSS france · west
Claude Sonnet 5
anthropic/claude-sonnet-5
Value destroyed$1,000Value lost$13,400 Units killed / lost10 / 5Buildings killed / lost0 / 8 Peak army$1,400Orders issued58 Decision turns (failed)37 (0)Mean latency4.9sModel cost$0.224
WIN england · east
GPT-5.6 Terra
openai/gpt-5.6-terra
Value destroyed$13,400Value lost$1,000 Units killed / lost5 / 10Buildings killed / lost8 / 0 Peak army$5,400Orders issued73 Decision turns (failed)37 (0)Mean latency3.4sModel cost$0.273

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-winnergpt56terra won: base destroyed
passcode_checkdecisions-loggedturns per slot: {'Multi0': 37, 'Multi1': 37}
passcode_checkllm-turns-validfailed/total per slot: {'Multi0': '0/37', 'Multi1': '0/37'}
passcode_checkvideo-decodesduration 50.3s
passcode_checkvideo-completevideo ends at tick 7248 of 7248
passcode_checklimits-respectedended after 7.3 wall-minutes (budget 30.0 min): base destroyed
passchecklistnotes-explain-winsummary.md Notes section cites both models final thoughts and explains Terra infantry push that destroyed Sonnet5 base
passchecklistplayers-match-requestresult.json Multi0=anthropic/claude-sonnet-5 (sonnet5) and Multi1=openai/gpt-5.6-terra (gpt56terra) match request
passchecklistno-silent-outagerun.log shows 37 consecutive turns for each player with 0 LLM failures and no gaps

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 →