Ollama: Weekly Recap - MLX Threading & Model Recommendations
Major MLX threading fixes for macOS stability, new experimental model recommendations endpoint, and significant batching improvements landed this week. The team also expanded launch integrations with Claude App support and improved Windows compatibility.
Duration: PT2M39S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-05-04T00:00:00Z
- Audio duration: PT2M39S
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 weekly recap for April 27th through May 4th, 2026.
16 PRs merged, 18 additional commits this week.
Starting with infrastructure improvements. The biggest change this week was a comprehensive MLX threading overhaul that addresses stability issues on macOS. MLX now uses OS-thread-local execution state, requiring all MLX operations to run on locked OS threads to prevent crashes from Go goroutine migration. This…
Metal GPU support also received hardening for initialization failures. The system now includes retry logic when tensor API compilation fails and better error handling to prevent crashes on systems with incomplete kernel coverage.
For model capabilities, significant batching groundwork landed with support for multi-sequence processing in the MLX runner. Models now use a Batch struct for forward inputs and support per-row positioning for RoPE operations, though single-sequence behavior remains unchanged. The attention cache system was also…
New features include an experimental model recommendations endpoint that replaces hardcoded model lists with server-driven recommendations. The desktop app now sources featured models from this endpoint,…
Nearby episodes from Ollama
- Claude Desktop Integration Removed
- Launch Command Enhancements
- Speed Revolution - MTP Decoding and Smart Caching
- Go 1.26 Runtime Update
- MLX Threading Fixes and Claude App Integration
- Model Recommendations and Windows Gateway Fix
- Metal GPU Stability and Gemma4 Updates
- Launch Experience Improvements and Model Recommendations