MMPCBench - Beyond Capability PCQ: 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. PCQ on questions whose answer cannot be inferred from the given evidence, such as unknowable futures; 0-100, higher is better.
Metric: PCQ (0-100). Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 93.9 | #280 |
| 2 | GPT-5.1 (High) | 87.5 | #145 (GPT-5.1) |
| 3 | Claude Sonnet 4.5 (Thinking) | 85.4 | #148 (Claude Sonnet 4.5) |
| 4 | Qwen 3 VL 8B Instruct | 84 | #419 |
| 5 | GPT-5.2 | 81.3 | #115 |
| 6 | Qwen 3 VL 30B A3B Instruct | 80.7 | #390 |
| 7 | GPT-5 Mini | 65 | #193 |
| 8 | Gemini 3 Pro | 64.2 | #82 |
| 9 | Gemini 3 Flash | 62.8 | #103 |
| 10 | GLM-4.6V | 54.6 | #330 |
| 11 | Gemma 3 12B (IT) | 48.2 | #672 |
| 12 | Gemma 3 27B (IT) | 43.4 | #528 |
| 13 | Gemma 3 4B (IT) | 22.4 | #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-beyond-capability-pcq · How It Works · Data refreshed daily, snapshot 2026-11-05.