SpecVQA: leaderboard
Metric: Accuracy (0-1, times 100), the unweighted mean of the English and Chinese descriptive and reasoning splits; expert-revised question-answer pairs on 620 spectra (NMR, IR, XRD, Raman, MS, UV-Vis, XPS) from published papers; each answer judged correct or not by GPT-o4-mini against the reference, numeric answers within a 5 percent tolerance; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 78.72 |
| 2 | Gemini 3 Pro (Preview) | 77.87 |
| 3 | Gemini 2.5 Pro | 76.15 |
| 4 | Gemini 2.5 Flash | 72.74 |
| 5 | GPT-5 (High) | 70.59 |
| 6 | O4 Mini (2025-04-16) | 70.54 |
| 7 | GPT-5 | 70.39 |
| 8 | O3 (2025-04-16) | 70.31 |
| 9 | GPT-5 (Low) | 69.61 |
| 10 | GPT-5.1 (Non-reasoning) | 65.78 |
| 11 | GPT-5.2 (Non-reasoning) | 65.4 |
| 12 | Claude Sonnet 4.5 (Thinking) | 59.41 |
| 13 | Claude Sonnet 4 (Thinking) | 56.88 |
| 14 | Qwen 3 VL 8B (Thinking) | 56.31 |
| 15 | Doubao-Seed-1.6 (Thinking) | 56.02 |
Interactive version: theaggregate.ai/benchmark?slug=specvqa · How It Works · Data refreshed daily, snapshot 2026-10-07.