Open-Weight Vision Benchmarks
Model Arena Leaderboard
Compare open-weight vision models head-to-head on the exact same product photos. Evaluate copywriting tone, latency, and tokens/sec.
Loading arena metrics...
Install Any 2 Local Vision Models
The arena automatically benchmarks any 2 installed Ollama vision models. Recommended models for 8GB+ RAM:
gemma3:4b (Default Model A)
Google open weights, ultra-efficient on 8GB RAM
llama3.2-vision (Default Model B)
Meta multimodal vision model, crisp detail detection
Recent Battle History (0)
No battle history found.