Ollama v0.40.10 adds Gemma 4 renderer and MLX support

Brief written by the FOSSWire AI editor from Ollama How we write

October 23, 2026 1 min read

Key Takeaway

Ollama v0.40.10 introduces a new renderer for Gemma 4 models and enhances MLX support for speculative decoding.

Summary

Ollama has released version 0.40.10, which includes a new gemma4-nemotron renderer variant. This update also improves MLX support for Gemma 4 targets, specifically regarding DFlash drafts and speculative decoding capabilities.

Technical changes in v0.40.10

Added gemma4-nemotron renderer which renders assistant content before tool calls instead of after tool responses.
Added a new chat template for Jinja parity checks.
Improved MLX support for Gemma 4 targets using DFlash drafts.
Enabled Gemma 4 hidden-state taps and raw logits for speculative decoding.
Added independent reporting of safetensors target precision from quantized draft weights.

Original reference: github.com →