MaCBench: leaderboard
Chemistry and materials multimodal benchmark evaluating VLMs across lab, molecular, crystallography, MOF, spectroscopy, patent-figure, table-QA, and XRD tasks.
Metric: Overall Score (score). Source: huggingface.co. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 4 Maverick Instruct | 0.7 | #456 |
| 2 | Claude 3.5 Sonnet | 0.67 | #350 |
| 3 | Llama 4 Scout Instruct | 0.63 | #539 |
| 4 | Gemini 1.5 Pro | 0.57 | #445 |
| 5 | GPT-4o | 0.54 | #343 |
| 6 | Mistral Small 3.1 | 0.53 | #610 |
| 7 | Grok 2 (1212) | 0.46 | #496 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=macbench · How It Works · Data refreshed daily, snapshot 2026-10-20.