Multimodal benchmark
SpatialViz-Bench leaderboard
As of October 2026, Gemini 2.5 Pro has the highest published SpatialViz-Bench score on Noometry at 44.7%, out of 8 models with results.
Last verified
About SpatialViz-Bench
Spatial visualization tasks such as mental rotation and paper folding.
- Category
- Multimodal
- Introduced
- 2025
- Unit
- Percent (random guessing ≈ 0%)
- Official site
- github.com
Top 8 models
- Gemini 2.5 Pro 44.7%
- o1 41.4%
- Gemini 2.5 Flash 36.9%
- Llama 4 Scout 34.2%
- Claude 3.7 Sonnet 33.9%
- Qwen2.5-VL 72B Instruct 33.3%
- Qwen-VL Max 32%
- Llama 4 Maverick 31.8%
Sponsored placements are available on pages like this one. Advertise on Noometry
All results
| # | Model | Provider | Score | Setting | Source | Date |
|---|---|---|---|---|---|---|
| 1 | Gemini 2.5 Pro |  Google | 44.7% | Epoch AI | ||
| 2 | o1 | OpenAI | 41.4% | Epoch AI | ||
| 3 | Gemini 2.5 Flash | Google | 36.9% | Epoch AI | ||
| 4 | Llama 4 Scout |  Meta | 34.2% | Epoch AI | ||
| 5 | Claude 3.7 Sonnet | Anthropic | 33.9% | Epoch AI | ||
| 6 | Qwen2.5-VL 72B Instruct |  Alibaba (Qwen) | 33.3% | Epoch AI | ||
| 7 | Qwen-VL Max | Alibaba (Qwen) | 32% | Epoch AI | ||
| 8 | Llama 4 Maverick | Meta | 31.8% | Epoch AI |
Compare the leaders
Other multimodal benchmarks
- LMArena Vision
- Video-MME
- GeoBench
- VPCT
- Blueprint-Bench 2
- Furniture Assembly
- LMArena Document (reference)
- MindCube (reference)
- ScienceQA (reference)
Frequently asked questions
What does SpatialViz-Bench measure?
Spatial visualization tasks such as mental rotation and paper folding.
Which model has the highest SpatialViz-Bench score?
As of October 2026, Gemini 2.5 Pro has the highest published SpatialViz-Bench score on Noometry at 44.7%, out of 8 models with results.
What is the best open-weight model on SpatialViz-Bench?
Llama 4 Scout has the highest SpatialViz-Bench accuracy among open-weight models at 34.2%, ranking 4 of 8 overall.