MMGist: leaderboard
We conduct a systematic study of 18 widely used vision-language benchmarks and identify three major issues: 1) many items do not rely on visual cues and therefore fail to effectively measure multimodal understanding; 2) many items are already close to performance saturation for current LVLMs, which limits their discriminative power; 3) a small number of anomalous items affect the reliability of ev
Metric: Macro ↑ (self-reported). Source: benchmarklist.com. Saturation forecast: Around April 2027. 27 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 66.8 | #58 |
| 2 | Qwen 3.6 Plus | 61 | #120 |
| 3 | Seed 2.0 Pro | 58.2 | #118 |
| 4 | Qwen 3.6 35B A3B | 55.2 | #226 |
| 5 | Gemini 3.1 Flash Lite | 52.7 | #191 |
| 6 | Doubao Seed 2.0 Mini | 52.4 | #184 |
| 7 | GPT-5 | 51.4 | #98 |
| 8 | Gemma 4 31B | 49.3 | #200 |
| 9 | Qwen 3.5 27B | 47.3 | #187 |
| 10 | Claude Sonnet 4.6 | 47.2 | #90 |
| 11 | Qwen 3.5 9B | 46.8 | #379 |
| 12 | Gemma 4 26B A4B | 45.6 | #358 |
| 13 | GPT-5 Mini | 45 | #186 |
| 14 | Step3 VL 10B | 43.3 | #485 |
| 15 | Qwen 3.5 4B | 42 | #475 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmgist · How It Works · Data refreshed daily, snapshot 2026-10-20.