Kimi-VL-A3B-Instruct: benchmark results
Provider: Moonshot. Released 2025-04-10. Access: Open.
Aggregate Rating 1511 ± 12, rank #2192 of 10916 rated models, from 714 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OmniHandwritingOCR - Easy Multi-Line Formulas | 88.82 | 1-NED (%; one minus the normalized Levenshtein distance betw | 100 |
| OmniHandwritingOCR - Single-Line Formulas | 93.6 | 1-NED (%; one minus the normalized Levenshtein distance betw | 100 |
| OpenVLM OCR-VQA | 82 | Accuracy (%) | 100 |
| OpenVLM TextVQA Val | 92.5 | VQA Score (%) | 100 |
| OpenVLM OCRBench - Handwritten Mathematical Expression Recognition | 92 | Correct Answers (%) | 99.8 |
| OpenVLM MMT-Bench - Referring Detection | 85 | Score (%) | 99.5 |
| OpenVLM ChartQA Test - Augmented | 95 | Relaxed Accuracy (%) | 99.1 |
| OpenVLM MMT-Bench - TU | 55.6 | Score (%) | 99 |
| OpenVLM MMT-Bench - Google Apps | 70 | Score (%) | 98.3 |
| OpenVLM MMT-Bench - Jigsaw Puzzle Solving | 45 | Score (%) | 98.3 |
| OpenVLM MMT-Bench - Musical Instrument Recognition | 100 | Score (%) | 98.3 |
| OpenVLM MMT-Bench - Salient Object Detection RGB | 85 | Score (%) | 98.3 |
Interactive version: theaggregate.ai/model?slug=kimi-vl-a3b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.