Sberbank Unveils GigaChat 3.5-432B Model

·2026.07.06 19:34

Key point

Sberbank has released the GigaChat 3.5 model with 432B parameters along with a GGUF quantized version.

Details

The GigaChat 3.5-432B-A28B model developed by Sberbank has been released on Hugging Face. The model supports the GGUF format alongside the Base version, improving inference efficiency in local environments.

Currently, the GGUF version has not yet been merged into the main branch of llama.cpp, but it can be built and used directly via the provided Pull Request.

llm gguf llama.cpp open-source-model gigachat sberbank

Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.