Multivac Evaluation: leaderboard
Historical Multivac tracker: mean positive peer-evaluation scores across each model's recorded rounds, with at least five positive-score evaluations required. The source excludes zero judgments before computing round scores; models have different category and judge panels. This is separate from the later paper's zero-inclusive, category-level analysis.
Metric: Avg Peer Score (0-10). Source: github.com. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seed 1.6 Flash | 9.45 | #436 |
| 2 | Grok 4.1 Fast | 9.18 | #217 |
| 3 | Grok Code Fast 1 | 9.12 | #337 |
| 4 | MiMo-V2-Flash | 8.96 | #372 |
| 5 | Claude Opus 4.5 | 8.95 | #83 |
| 6 | Claude Sonnet 4.5 | 8.92 | #142 |
| 7 | Gemini 3 Flash (Preview) | 8.92 | #82 |
| 8 | Gemini 2.5 Flash | 8.72 | #245 |
| 9 | Gemini 2.5 Flash Lite | 8.72 | #422 |
| 10 | GPT-5.2 Codex | 8.7 | #95 |
| 11 | GPT-OSS-120B | 8.58 | #341 |
| 12 | GLM-4.7 | 7.28 | #192 |
| 13 | OLMo 3.1 32B (Thinking) | 7.23 | #632 (OLMo 3.1 32B) |
| 14 | Gemini 3 Pro (Preview) | 7.2 | #66 |
Interactive version: theaggregate.ai/benchmark?slug=multivac-evaluation · How It Works · Data refreshed daily, snapshot 2026-10-20.