Ollama: Stricter Model Imports and CLI Output Fixes
The dominant pattern is correctness, led by unified expert stacking and lossless eight-bit import handling, alongside tighter version, wrapping, byte formatting, and port preservation fixes.
Duration: PT2M16S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-11-09T14:01:28Z
- Audio duration: PT2M16S
Transcript excerpt
This excerpt keeps the crawler page concise. Listen to the episode or use the RSS feed for the full update.
It's November 9th, 2026, and this is your Ollama developer briefing.
The throughline today is correctness over convenience — making imports stricter and command-line output more honest. No single outage drives the day; these are independent reliability wins.
First, model creation gets more consistent. In pull request 19614, expert stacking for mixture-of-experts models now uses the same planner regardless of source, fixing cases where some models shipped with unstacked experts. On the precision side, per-tensor eight-bit float now imports exactly, while block and…
Second, a cluster of small fixes tightens what you see in the terminal. Pull request 19618 stops warning when you simply ask for the version with no server running — it now just prints the installed client version. Pull request 19620 fixes line wrapping for complex emoji, counting a multi-character glyph once…
Finally, configuration and docs get clearer. Pull request 19619 preserves a custom port when network exposure switches the bind address to all interfaces. Pull request 19612 documents how to discover embedding size when the dimensions option is omitted, linking the show model info field to the embeddings response…
What…
Nearby episodes from Ollama
- Tool Reliability and Mac Inference Tuning
- Faster Concurrent Inference and Unified Desktop Launch
- MLX Goes Parallel
- VRAM Fixes and Launch Polish
- Weekly Recap - MLX Model Expansion and Launch Polish
- Ripple on MLX and Validation Hardening
- MLX Memory Cuts and Tool-Call Fixes
- Launcher Return, API Fixes and Inference Reliability