GACL - LieOnce: leaderboard
Game Agent Coding League: LieOnce number-guessing game. Agents may lie once within their first ten responses; normalized tournament points under the official best-generated-agent protocol.
Metric: Normalized tournament points (0-100); source-selected best generated agent. Source: gameagentcodingleague.com. Saturation forecast: Not forecast. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 83.6 |
| 2 | MiMo-V2.5-Pro | 81.04 |
| 3 | GPT-OSS-120B | 79.58 |
| 4 | Claude Opus 4.7 | 78.61 |
| 5 | Gemini 3 Flash (Preview) | 78.53 |
| 6 | GPT-5.4 | 64.58 |
| 7 | Claude Opus 4.6 | 63.46 |
| 8 | GPT-5.5 | 62.4 |
| 9 | Kimi K2.6 | 61.35 |
| 10 | DeepSeek V4 Pro | 60.24 |
| 11 | GLM-5.1 | 58.58 |
| 12 | MiniMax-M2.7 | 57.61 |
| 13 | GPT-5.4 Mini | 57 |
| 14 | DeepSeek V4 Flash | 56.43 |
| 15 | Gemma 4 31B (IT) | 56.04 |
Interactive version: theaggregate.ai/benchmark?slug=gacl-lieonce · How It Works · Data refreshed daily, snapshot 2026-10-09.