Overview

Gemma 3 is Google DeepMind’s third generation of open-weight models built from the same research as Gemini. Released in March 2025 in 1B, 4B, 12B and 27B sizes, it added image understanding to the larger models and extended the context window, while staying small enough to run on a single GPU or on devices.

Highlights

  • Four sizes: 1B, 4B, 12B and 27B parameters
  • Image input on the 4B, 12B and 27B models
  • 128K-token context window on larger sizes
  • Support for over 140 languages
  • Quantized versions for efficient local use

Typical uses

Local assistants, on-device applications, fine-tuning for specialized tasks and multilingual products.

Things to consider

  • Gemma uses its own terms of use with a prohibited-use policy, not an OSI open-source license.
  • The 1B model is text-only with a shorter context.
  • Newer Gemma versions may be available.

Specifications are taken from the provider’s official announcement, documentation or model card. Benchmark results depend on evaluation settings; see how AI benchmarks work.