C.W.K.
Stream
Lesson 01 of 04 · published

Vision-Capable Local Models

~18 min · vision, multimodal

Level 0Downloader
0 XP0/41 lessons0/11 achievements
0/120 XP to next level120 XP to go0% complete

Local vision in 2026

Vision models accept images alongside text and answer questions about them. The 2026 local landscape is good enough for OCR, chart reading, screenshot analysis, and most "describe this image" tasks. It's not yet at frontier quality (GPT-4V, Claude 3.5 Sonnet vision, Gemini 3) for hardest reasoning tasks.

ModelSizesStrength
Llama 3.2-Vision11B, 90BStrong general vision; well-supported in Ollama
Gemma 34B / 12B / 27BBest quality-per-size; runs on modest hardware
Qwen 2.5-VL3B / 7B / 32B / 72BBest for structured data extraction, OCR, multilingual
Qwen 3.5variousVision + tools + thinking — most capable single model
LLaVA7B / 13B / 34BThe OG open vision model; still serviceable
MiniCPM-V8BCompact, capable, good for edge deployment
Mistral Small 3.1variousVision support added in 3.1

Picking by use case

  • OCR / structured extraction: Qwen 2.5-VL is the strongest open option.
  • General "describe this": Gemma 3 12B/27B or Llama 3.2-Vision 11B.
  • Tight memory: Gemma 3 4B or MiniCPM-V 8B.
  • Vision + tools + thinking: Qwen 3.5 (when available in Ollama for your size).

Always verify capability

Vision support is binary at the model level — either the model has the visual encoder or it doesn't. ollama show MODEL capabilities line is the source of truth.

Code

Verify and pull vision models·bash
# Check capability (look for 'vision' in capabilities)
ollama show llama3.2-vision | grep -i capabilities
ollama show gemma3:12b      | grep -i capabilities
ollama show qwen2.5vl       | grep -i capabilities

# Pull a daily-driver vision model
ollama pull gemma3:12b

# Pull an OCR-strong model
ollama pull qwen2.5vl:7b

External links

Exercise

Pull two vision models (one general, one OCR-strong) and run ollama show capabilities on each. Hit each with the same image and the same prompt; note which one wins on the task you actually care about.

Progress

Progress is local-only — sign in to sync across devices.
Spotted a bug or have feedback on this page?Report an Issue

Comments 0

🔔 Reply notifications (sign in)
Sign inPlease sign in to comment.

No comments yet — be the first.