Ollama Now Supports Gemma 3 — Here's What That Means

Google's Gemma 3 landed on Ollama this week, and it's worth your attention.

Why Gemma 3 matters

Most 7B models feel like toys. Gemma 3 at 7B parameters punches above its weight class — rivaling 13B models from six months ago. The 128K context window is genuinely usable on consumer hardware.

What you need to know

SpecValue

Parameters7B Context128K tokens VRAM~6GB at Q4 Speed~40 tok/s on RTX 4090

How to run it

`bash ollama pull gemma3:7b ollama run gemma3:7b `

That's it. One command.

Where it shines

RAG pipelines: The 128K context means you can stuff entire documents in without chunking

Summarization: Clean, structured outputs without much prompting

Code explanation: Not as good as dedicated code models, but solid for understanding unfamiliar codebases

Where it falls short

Creative writing: It's noticeably worse than Llama 3 at narrative tasks

Math: Simple arithmetic is fine; multi-step reasoning struggles

Instruction following: Can be overly literal, missing implied intent

Bottom line

Gemma 3 is the best small model you can run on consumer hardware right now. It won't replace GPT-4o or Claude for complex tasks, but for local RAG and everyday Q&A, it's the new default.

Set up Ollama →