Kimi-VL-A3B-Thinking-2506: benchmark results
Provider: Moonshot. Released 2025-06-21. Access: Open.
Aggregate Rating 1561 ± 19, rank #1525 of 10916 rated models, from 187 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OpenVLM MathVista - ARI | 76.8 | Accuracy (%) | 99.7 |
| OpenVLM MathVista - STA | 89.4 | Accuracy (%) | 99.1 |
| OpenVLM MathVista - VQA | 65.4 | Accuracy (%) | 98.1 |
| OpenVLM MathVista - FQA | 79.9 | Accuracy (%) | 97.6 |
| OpenVLM MMStar - Instance Reasoning | 78.4 | Accuracy (%) | 97 |
| OpenVLM MathVista - MWP | 88.7 | Accuracy (%) | 96.9 |
| OpenVLM MathVista - GEO | 85.4 | Accuracy (%) | 96.8 |
| OpenVLM MathVista - GPS | 87 | Accuracy (%) | 96.6 |
| TimeSpot - Local Time | 32.23 | Accuracy (%) of the predicted local time of day within one h | 96.2 |
| OpenVLM MMMU - Art Theory | 93.3 | Accuracy (%) | 96.1 |
| OpenVLM MathVista - ALG | 83.3 | Accuracy (%) | 96.1 |
| MathVista | 79.5 | Accuracy (%) | 95.8 |
Interactive version: theaggregate.ai/model?slug=kimi-vl-a3b-thinking-2506 · How It Works · Data refreshed daily, snapshot 2026-11-08.