MCSBench - Overall: leaderboard
Measures multimodal conceptual-structure alignment on MCSBench, a diagnostic multiple-choice benchmark of 661 base visual question-answering items over fine-grained visual and taxonomy-like categories, spanning seven task types and mixing single- and multi-image records. Overall is accuracy across all 661 base questions. These are base-VQA accuracies from the released results table; the reasoning-chain integrity overlay and its Structural Alignment Index are scored separately and are not part of this series. Five API models also have a run with web search enabled in that table; those runs are not part of this series either. This series reports MCSBench - Overall using Accuracy (%).
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Around February 2027. 77 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro (Preview) | 82.6 | #71 |
| 2 | Gemini 3 Flash (Preview) | 80.79 | #90 |
| 3 | GPT-5 | 78.37 | #105 |
| 4 | Claude Opus 4.6 | 75.19 | #67 |
| 5 | Gemini 2.5 Pro | 74.13 | #159 |
| 6 | Gemini 2.5 Flash | 66.57 | #255 |
| 7 | O4 Mini | 66.26 | #185 |
| 8 | GPT-5 Mini | 63.69 | #193 |
| 9 | Qwen 3 VL 235B A22B (Thinking) | 62.63 | #236 |
| 10 | Qwen 3 VL 235B A22B Instruct | 62.18 | #280 |
| 11 | Qwen 2.5 VL 72B Instruct | 60.06 | #383 |
| 12 | Qwen 3 VL 30B A3B (Thinking) | 59.15 | #348 |
| 13 | Claude Sonnet 4.5 | 59 | #148 |
| 14 | Qwen 3 VL 30B A3B Instruct | 58.85 | #390 |
| 15 | GLM-4.5V | 58.09 | #358 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mcsbench-overall · How It Works · Data refreshed daily, snapshot 2026-11-05.