Ollama: Gemma 4 Support and Platform Improvements

Ollama's June 15th activity centers on comprehensive Gemma 4 model support with new MLX architecture implementations and memory optimizations, alongside critical fixes for multimodal processing and developer tooling improvements.

Duration: PT2M24S

Episode overview

This episode is a short developer briefing from Ollama.

It explains recent repository work in plain language.

  • Show: Ollama
  • Published: 2026-06-15T13:01:10Z
  • Audio duration: PT2M24S

Transcript excerpt

This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.

Good morning, this is your Ollama developer briefing for June 15th, 2026.

The dominant theme today is comprehensive Gemma 4 support across Ollama's infrastructure. Three major pull requests introduce MLX runner support for both standard Gemma 4 models and the new diffusion-gemma architecture, with PR 16741 addressing mixed-precision quantization and PR 16743 implementing block-diffusion…

Memory management improvements extend beyond Gemma 4. PR 16742 introduces MLX memory capping to prevent runner crashes on memory-constrained Macs, while PR 16745 replaces the blanket 10 gigabyte multimodal projector offload with size-based calculations. These changes directly address user reports of models becoming…

Developer tooling sees several important fixes. The file path extraction system gets two critical repairs in PRs 16735 and 16729, preventing incorrect splitting of paths when parent directories contain image extensions. PR 16730 preserves UNC network path prefixes on Windows. The progress display system gets a…

Additional updates include llama.cpp advancing to version b9637, optional thinking support for LFM 2 models, and new AMD EPYC CPU acceleration through ZenDNN backend support.

These…

Nearby episodes from Ollama

  1. Weekly Recap - MLX Performance & Path Handling
  2. Memory Management and Multimodal Parsing Fixes
  3. GPU Offloading and Tool Call Fixes
  4. Performance Optimizations and Model Handling Improvements
  5. Infrastructure Updates and Platform Fixes
  6. Multimodal Fixes and Developer Experience Updates
  7. Cache Architecture Overhaul and Data Race Fixes
  8. Developer Tools and Cross-Platform Reliability