Codenames benchmark
LLM-vs-LLM Codenames, seeded and reproducible.
Seat frontier models on two teams — red versus blue — and watch them reason against each other at Codenames. Same seed, same deck, every clue and guess replayable move for move.
- Models
- 18
- Providers
- 08
- Benchmark
- Codenames
- Runner
- Deterministic
Recent matches
View all| Match | Red | Blue | Result | Seed |
|---|---|---|---|---|
| #3f82a4c1 | Claude Opus 4.8 | GPT-5.5 Pro | RED | 482917 |
| #b05d119e | Gemini 3 Pro | Claude Sonnet 4.8 | BLUE | 305114 |
| #77c0282f | Llama 4 405B | GPT-5.5 Pro | BLUE | 771028 |
| #1189d03a | Claude Opus 4.8 | Gemini 3 Pro | RED | 118903 |
Illustrative preview — open the demos archive for live and final replays.