Ollama default
Download: 8.2 GB · Estimated planning budget: ~11.1 GB
ollama run gemma3:12bVerify on Ollama Source reviewed 2026-10-14
Google · Open weights
Multimodal assistant with image-and-text support.
Model size
12B
Reported license
Gemma license
Approximate 4-bit GPU starting point
12 GB+
Variant-specific artifact sizes and runtime tags are sourced from Ollama. Planning budgets are N44 estimates, not benchmarks or tested requirements.
Download: 8.2 GB · Estimated planning budget: ~11.1 GB
ollama run gemma3:12bVerify on Ollama Source reviewed 2026-10-14
The 12 GB figure is only an editorial starting estimate for a selected low-bit inference configuration. Exact RAM and VRAM needs depend on the quantized file, context length, KV cache, runtime, and CPU offloading. It is not a tested minimum.
Check runtime support and the actual model format before downloading.
Review the publisher's license, model card, intended use, and limitations. Access approval may be required for gated repositories. N44's editorial summary is not a substitute for those terms.
Only results with documented sources, model revisions and evaluation settings appear here. Different benchmarks cannot be combined into a single reliable score.
No source-reviewed benchmark records are available for this model yet. This is not a zero score.
Visit the official publisher repository to review files, versions, access requirements and downloads. The original model files may require conversion or a supported quantization before running locally.
Official model repository