Ollama: LFM3 Speech, Video and Reliability Fixes
LFM3 work dominates, with GGUF correctness and text-to-speech on the server runner and video input on Mac MLX models. Smaller fixes improve download resume, networking, GPU selection, and CLI behavior.
Duration: PT2M35S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-10-28T13:01:47Z
- Audio duration: PT2M35S
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 October 28th, 2026 — you're listening to Ollama.
The through-line today is multimodal model readiness, led by L-F-M-3 fixes for speech and video, paired with a set of small reliability corrections for downloads, networking, and the command line.
First, L-F-M-3 across both runners. Merged work in PR 19344 corrects how the llama server runner handles those GGUF files — honoring an explicit beginning-of-sequence token setting, which cut WikiText perplexity on L-F-M-2 24-b from 81 to 72, mapping repeat-last-n to context size, and stopping audio-only models from…
Second, correctness for everyday flows. PR 19339 would keep interrupted downloads resumable across restarts by fixing startup cleanup that deleted stale part files and caused missing-file or digest errors. PR 19351 allows public I-P version 4 destinations through Nat 64 translation while still blocking private and…
Finally, four community integration listings — Paw Pulse monitoring, Voltip dictation, A-I-Term terminal, and interviewer A-I — with no behavior change.
What's next: watch for the llama dot cpp bump that actually switches on G-G-U-F speech, and for decisions on video support beyond Mac and non-MLX…
Nearby episodes from Ollama
- Sampler Defaults Fix and Memory Work
- MLX Memory Safety and Silent Correctness Fixes
- MLX Performance Push and Compatibility Fixes
- OpenAI Fidelity and Runner Stability
- Scheduler Memory Fixes and Faster Drafting
- Mac Prefill Gains and Scheduler Fixes
- Weekly Recap - Phonon Voice Goes GA and MLX Speed Gains
- Gemma 4 Reliability and Runner Parity