MMPCBench - Strategic Effectiveness: leaderboard
MMPCBench (arXiv 2608.29286v1, Table 2): proactive critique of flawed multimodal inputs for 14 MLLMs; a judge panel of GPT-5.2, Gemini-3-Pro and Claude-Sonnet-4 scores the responses, and two of those judges are also evaluated. Mean judged utility of the chosen correction, clarification or refusal on detected errors only, 0 to 2; higher is better.
Metric: Judge-panel score (0-2; detected errors only). Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.1 (High) | 1.95 | #145 (GPT-5.1) |
| 2 | GPT-5 Mini | 1.93 | #193 |
| 3 | GPT-5.2 | 1.87 | #115 |
| 4 | Claude Sonnet 4.5 (Thinking) | 1.86 | #148 (Claude Sonnet 4.5) |
| 5 | Gemini 3 Pro | 1.79 | #82 |
| 6 | Gemini 3 Flash | 1.77 | #103 |
| 7 | Qwen 3 VL 235B A22B Instruct | 1.69 | #280 |
| 8 | GLM-4.6V | 1.69 | #330 |
| 9 | Qwen 3 VL 30B A3B Instruct | 1.57 | #390 |
| 10 | Qwen 3 VL 8B Instruct | 1.54 | #419 |
| 11 | Gemma 3 27B (IT) | 1.42 | #528 |
| 12 | Gemma 3 12B (IT) | 1.42 | #672 |
| 13 | Gemma 3 4B (IT) | 1.24 | #997 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmpcbench-strategic-effectiveness · How It Works · Data refreshed daily, snapshot 2026-11-05.