Ollama v0.31.1 improves Gemma 4 MoE model loading

Ollama has released version 0.31.1, which includes improvements to the loading of Gemma 4 Mixture of Experts (MoE) models. This update allows for more flexible loading of both quantized and non-quantized versions of these models by standardizing tensor naming conventions. AI

IMPACT Enhances the usability and flexibility of running advanced AI models like Gemma 4 MoE on local hardware.

RANK_REASON This is a software release for a tool that facilitates running AI models locally, not a core AI model release or research.

Read on Ollama — Releases →

• Gemma 4
• Ollama

• infra
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →