Unsloth Optimizes Gemma 3: Google's Multimodal AI Now Faster, More Efficient; Supports 1B-27B Models with 128K Context

Fine-tune Gemma 3 with Unsloth

Fine-tune & Run Gemma 3Mar 10, 2025 • By Daniel & MichaelMar 14, 2025•By Daniel & MichaelGemma 3, Google's new state-of-the-art multimodal (text + image) models come in 1B, 4B, 12B, and 27B sizes. Now supported in Unsloth, Gemma 3 has a 128K context window, and multilingual support.Update March 19: Read about our new Gemma 3 training fixes hereFine-tune Gemma 3 for free using our Colab notebook. Also, view our Gemma 3 (1B) GRPO notebookUnsloth makes Gemma 3 (12B) finetuning 1.6x faster, use 60% ...