Ollama: Weekly Recap - Parallel Inference, Tool Fidelity, and Launcher Polish

Ollama merged parallel generation for MLX, reliability fixes for integrated GPUs, and safer tool schema handling alongside launcher improvements. The pattern this week was making local runtimes more concurrent, compatible, and predictable.

Duration: PT3M15S

Episode overview

This episode is a short developer briefing from Ollama.

It explains recent repository work in plain language.

  • Show: Ollama
  • Published: 2026-11-16T09:07:37Z
  • Audio duration: PT3M15S

Transcript excerpt

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

Hello and welcome to your Ollama briefing for November 9th through November 16th.

30 pull request activity items, 13 additional commits this week.

The throughline this week is making local models behave more like production runtimes: Apple silicon now handles parallel requests, integrated graphics loads more reliably, and tool definitions survive round-trips intact.

First, inference performance and hardware reliability. Merged work in pull request 19197 lets MLX models serve multiple generation requests at once, sharing projection work while keeping attention, caches and sampling isolated. The default remains a single request, with higher parallelism trading memory for faster…

Second, tool calling and reasoning controls. Pull request 19686 preserves unmodeled tool schema keywords like default and minimum, which were previously dropped when decoding and re-encoding parameters. Pull request 19578 adds support for deferred tool loading in Claude Code. Pull request 19684 maps a request to…

Third, launcher workflow and documentation. Pull request 19583 makes ChatGPT restore ask before changing settings and improves graceful restart with cancellation support. Pull request 19409 builds…

Nearby episodes from Ollama

  1. Tool Call Reliability and Runner Efficiency
  2. Runner Reuse, Decision Models, and Thinking UI Fixes
  3. MLX Concurrency and Memory Tuning
  4. GPU Load Fixes and Clearer Errors
  5. Qwen Tool-Call Reliability Fix
  6. Tool Schema Fidelity, Desktop Launch, and Download Safety
  7. Tool Reliability and Mac Inference Tuning
  8. Faster Concurrent Inference and Unified Desktop Launch