Ollama: Mac Polish, MLX Reliability, and Compatibility Fixes

Desktop work speeds up Settings and fixes macOS install prompts and update UI, while open work targets MLX upgrades and OpenAI data-loss bugs. The pattern is fewer freezes, false prompts, and silently dropped model output.

Duration: PT2M36S

Episode overview

This episode is a short developer briefing from Ollama.

It explains recent repository work in plain language.

  • Show: Ollama
  • Published: 2026-09-25T13:01:36Z
  • Audio duration: PT2M36S

Transcript excerpt

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

Good morning, it's September 25th, 2026, and this is your Ollama developer briefing.

The through-line today is reliability on the Mac and in the compatibility layer — fewer freezes, fewer false prompts, and fewer cases where model output gets silently mangled.

First, desktop polish. Three separate proposals — PRs 18645, 18643 and 18629 — all converge on the same bug: launching Ollama from a subdirectory of Applications, like Applications slash AI, incorrectly triggered a Move to Applications prompt, while similarly named apps at the top level slipped through. The fix…

Second, stop dropping developer data. On the OpenAI compatibility endpoint, PR 18635 adds reasoning content as an alias so reasoning text replayed from DeepSeek-style clients isn't lost, and PR 18639 preserves tool call IDs when tool results arrive as content parts. The same theme shows up in model parsers: PR 18624…

Finally, MLX on Mac is maturing. PR 18634 ensures older installs automatically pull the MLX variant when a tag becomes a manifest list, PR 18625 bounds stalled download retries and lets the watchdog actually interrupt them, and PR 18631 teaches the Gemma 4 loader to handle pre-stacked expert weights.

What's…

Nearby episodes from Ollama

  1. GPU Efficiency Fix and MLX Hardening
  2. Single-Pass Structured Outputs and MLX Speedups
  3. Agent Reliability Fixes and Offline Model Transfer
  4. Thinking Separation Fix and GPU Visibility
  5. Weekly Recap - MLX Graduation, Reasoning Fixes, and Onboarding Polish
  6. Compatibility Fixes and Mac Efficiency
  7. Closing the Gaps Between Model Metadata and Client Compatibility
  8. Hardening Pass Across Network, Memory, and Release Paths