Hugging Face's logo

Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
Back to feed
danielhanchen 
posted an update 3 days ago
eshucode

can you pls link a blogpost or something, what's the source for this post?

·
danielhanchen

It was posted officially by Google: https://x.com/googlegemma/status/2077449152062247219
plz12345

Only the chat templates were updated? Ex., checked the 26B A4B 8-bit MLX quant. I can't seem to reconcile "model is updated" with "last modified date 4 months ago".

·
danielhanchen

Our MLX quants were update: https://huggingface.co/collections/unsloth/gemma-4
dipankarsarkar

Speed is the headline. plz12345 already found the real story underneath: same hashes, only the template moved.

If the weights are byte-identical and just the chat template changed, that is still a real fix. The template is what decides whether function-call args parse. But it means every quant and the bf16 should improve together, not one more than another.

So the receipt that matters isn't 're-download the GGUF'. It is the same weights, old template vs new, tool-call pass rate before and after.

On one quant, how much did tool-call accuracy actually move between the two templates?

·
danielhanchen

If you read our graphic, it says you can update the template as well. Most people don't know how to replace the chat template.
In this post