Best Local Vision Models by VRAM (October 2026)
Open-weight vision-language models sized by VRAM: Qwen3-VL 2B to 235B-A22B, Mistral Small 3.2, Gemma 3, Pixtral, and where each tier breaks.
Tag
#vision
MiniMax-M3 Lands in llama.cpp: Sparse Attention and Vision
Two back-to-back merges add MiniMax Sparse Attention and a Qwen2.5-VL style vision tower to llama.cpp, but every existing MiniMax-M3 GGUF must be regenerated.
Best Local Vision Models: Every GPU Tier (August 2026)
Local image analysis, OCR, and visual reasoning from 8GB to 32GB VRAM. Qwen3.5 replaces Qwen3-VL at most tiers, and 16GB stays unresolved.