Google is updating Gemma 4's chat templates, bringing major fixes to tool calling and reducing "laziness", and enabling Flash Attention 4 on Hopper GPUs, plus an interactive guide on how to work with and improve its vision!
Summary
Google updated Gemma 4's chat templates with major fixes to tool calling, reduced laziness, enabled Flash Attention 4 on Hopper GPUs, and released an interactive vision guide. The updates are available on Hugging Face.
View Cached Full Text
Cached at: 07/15/26, 07:57 PM
A huge shoutout to the community for submitting fixes and finding new ways to make Gemma even better. We couldn’t do this without you! ❤️
🤗 Ready to test the speedup? Download the latest Gemma 4 updates now on Hugging Face: https://t.co/zAtd2yx7FB (5/5)
Gemma 4 - a google Collection
Source: https://huggingface.co/collections/google/gemma-4
- —
#### google/gemma-4-31B-it Image-Text-to-Text• 33B• Updatedabout 3 hours ago • 12.3M • • 3.21k - — #### google/gemma-4-31B Image-Text-to-Text• 33B• Updatedabout 3 hours ago • 817k • 465
- — #### google/gemma-4-26B-A4B-it Image-Text-to-Text• 27B• Updatedabout 3 hours ago • 14.2M • • 1.27k
- — #### google/gemma-4-26B-A4B Image-Text-to-Text• 27B• Updatedabout 3 hours ago • 1.06M • 341
- — #### google/gemma-4-E4B-it Any-to-Any• 8B• Updatedabout 3 hours ago • 5.77M • 1.37k
- — #### google/gemma-4-E4B Any-to-Any• 8B• Updatedabout 3 hours ago • 466k • 357
- — #### google/gemma-4-E2B-it Any-to-Any• 5B• Updatedabout 3 hours ago • 2.78M • 819
- — #### google/gemma-4-E2B Any-to-Any• 5B• Updatedabout 3 hours ago • 111k • 393
- — #### google/gemma-4-E2B-it-assistant Any-to-Any• 78M• Updatedabout 3 hours ago • 86.2k • 67
- — #### google/gemma-4-E4B-it-assistant Any-to-Any• 78.8M• Updatedabout 3 hours ago • 262k • 115
- — #### google/gemma-4-26B-A4B-it-assistant Any-to-Any• 0.4B• Updatedabout 3 hours ago • 285k • 169
- — #### google/gemma-4-31B-it-assistant Any-to-Any• 0.5B• Updatedabout 3 hours ago • 1.33M • 312
- — #### google/gemma-4-12B-it-assistant Any-to-Any• 0.4B• Updatedabout 3 hours ago • 99.3k • 100
- — #### google/gemma-4-12B-it Any-to-Any• 12B• Updatedabout 3 hours ago • 2.84M • 1.3k
- — #### google/gemma-4-12B Any-to-Any• 12B• Updatedabout 3 hours ago • 305k • 656
Similar Articles
Gemma 4 Chat Template now has preserve thinking
Google's Gemma 4 31B IT model now has a chat template fix that preserves thinking and improves null handling, reasoning preservation, and input validation.
@googlegemma: We’re rolling out some big improvements to Gemma 4, fueled by incredible community feedback and contributions! Here is …
Google Gemma is rolling out significant improvements to Gemma 4, driven by community feedback and contributions, as detailed in a thread.
Google AI Edge Gallery v1.0.13 & v1.0.14 updates: Gemma 4 Multi-Token Prediction, Pixel TPU support, experimental MCP, new skills, now saves chat history
Google AI Edge Gallery v1.0.13 & v1.0.14 updates add support for Gemma 4 with multi-token prediction, Pixel TPU optimization, experimental MCP, new skills, and chat history saving, enhancing on-device generative AI capabilities.
Welcome Gemma 4: Frontier multimodal intelligence on device
Google DeepMind releases Gemma 4, a frontier multimodal model family available on Hugging Face with Apache 2 licensing, optimized for on-device deployment and supported by various inference libraries.
With Gemini 3.5 Flash, Google bets its next AI wave on agents, not chatbots
Google launched Gemini 3.5 Flash, a new AI model optimized for coding and autonomous agents, shifting focus from chatbots to agentic AI. It outperforms previous models and powers new products like Antigravity 2.0 and Gemini Spark.