作者： Jakub Rusinowski · 最后更新： 2025年3月12日
Google最强大的开源模型系列。Gemma 3提供出色的VRAM性价比，4B模型超越许多7B竞品。具备多模态能力，所有规格均支持128k上下文窗口。
| Licence | What it permits | Applies to |
|---|---|---|
Gemma Terms | Commercial use permitted Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms. | Gemma 3 1B Instruct, Gemma 3 4B Instruct, Gemma 3 12B Instruct, Gemma 3 27B Instruct |
| Gemma 3 1B Instruct | Min 1 GB VRAM · Q4_K_M · 32,000 ctx · ollama run gemma3:1b |
| Gemma 3 4B Instruct | Min 3 GB VRAM · Q4_K_M · 128,000 ctx · ollama run gemma3:4b |
| Gemma 3 12B Instruct | Min 8 GB VRAM · Q4_K_M · 128,000 ctx · ollama run gemma3:12b |
| Gemma 3 27B Instruct | Min 17 GB VRAM · Q4_K_M · 128,000 ctx · ollama run gemma3:27b |
The cheapest GPU that runs Gemma 3 locally (min 1 GB VRAM) is the Intel Arc B570 (10 GB).
Install Ollama then run: ollama run gemma3:1b
Minimum VRAM: 1 GB. For best results use Q4_K_M quantization.
Gemma 3 needs about 1 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Gemma 3 1B Instruct (1 GB, Q4_K_M); Gemma 3 4B Instruct (3 GB, Q4_K_M); Gemma 3 12B Instruct (8 GB, Q4_K_M); Gemma 3 27B Instruct (17 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Yes — Gemma 3 runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
Q4_K_M is the best balance of quality and VRAM for Gemma 3 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
Install Ollama, then run: ollama run gemma3:1b. This downloads Gemma 3 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.