← Every modelGoogle DeepMind · 27B, dense

Gemma 3 27B

Google's largest Gemma 3. Reads images, writes well, and handles many languages. It needs a 24 GB card to run with room for context.

ollama run gemma3:27b

Needs about 19 GB of graphics memory, or unified memory on a Mac. First time? Install Ollama, then paste that into a terminal.

Download17 GB
Memory to runAbout 19 GB
Read per token17 GB
LicenseGemma Terms of Use
Context128K tokens
ReleasedMarch 2025
Good for
  • Image understanding with a smarter model
  • Polished writing and translation
  • Long documents
Not the best pick for
  • Speed: as a dense 27B it is slower than the MoE models of similar size
63 GPUs and Macs

What runs Gemma 3 27B

Estimated speed on each card at the smallest memory size that fits it, from memory bandwidth. As a dense model it reads all its weights for every token, so memory bandwidth sets the speed.

Runs great 14 cards

Usable 32 cards

Slow 17 cards

Cheapest build from our list that runs it well: Used workstation + 2× Tesla P100, from $750 used, at about 24 tok/s. See the builds.

Learn

Read more about Gemma 3 27B

Compare

Other models