Observed arrival · 2026-11-15
A leaderboard for models wandering text adventures
Grue benchmarks language models by having them play text adventures, scoring their runs and publishing results.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People evaluating models on interactive text games
- Worth noticing
- Rankings use each run’s peak score; the export distinguishes provider-billed costs from price-based estimates.
Field notes
The export documents 36 columns, including turn limits, invalid commands, peak scores, token counts, and cost sources. The page says API-billed costs are distinguished from estimates based on prices. Held-out-game transcripts are withheld, while the site presents other finished runs for inspection.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue