4 observations · 2 reports · 2 publishers
Win Rate.
Every Win Rate observation from our practitioner benchmark corpus. Latest: Unknown publication (2026).
Observations
4
Reports
2
Publishers
2
Latest
Unknown publication
2026-02-23
2026-02-23
62.0 %
Unknown publication
2026
Gemini 3 Pro highest
Model-vs-model Tetris matches (800+ games)
TetrisBench: How Language Models Develop Playing Strategies Through Code Generation →
Model-vs-model Tetris matches (800+ games)
TetrisBench: How Language Models Develop Playing Strategies Through Code Generation →
60.3 %
Unknown publication
2026
Gemini 3 Flash
Model-vs-model Tetris matches (800+ games)
TetrisBench: How Language Models Develop Playing Strategies Through Code Generation →
Model-vs-model Tetris matches (800+ games)
TetrisBench: How Language Models Develop Playing Strategies Through Code Generation →
14 percent
cdn.prod.website-files.com
2025
28 percent
cdn.prod.website-files.com
2025