Ollama: Gemma 4 Reliability and Runner Parity
Three Gemma 4 fixes address vision detail, dropped tool calls, and blank model metadata, alongside work to bring the MLX runner to parity on stop sequences and improve memory reporting and long-session compaction.
Duration: PT2M15S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-10-25T13:00:39Z
- Audio duration: PT2M15S
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 25th, 2026 — you're listening to the Ollama daily briefing.
The big pattern is reliability for compact models and alternate runners, with three separate fixes aimed at making Gemma 4 usable in real agent workflows.
First, Gemma 4. The compact vision setup was under-budgeting small images, dropping detail below its intended token budget. That fix merged yesterday as pull request 19247. A second issue hits tool calling: when the small Gemma variant leaves a list unclosed in long edit arguments, the parser dropped the call…
Second, parity and resilience between engines. Stop sequences were reaching the MLX runner but never applied, so the same prompt would stop correctly on the llama path and run on until the end on MLX. Pull request 19267 brings them in line, which matters for voice assistants and raw-prompt clients that rely on…
Finally, smaller housekeeping: a user interface fix stops the sidebar from animating open on every load, and three readme updates refresh community links and integrations.
What's next: if you run Gemma 4 or MLX models, retest tool calls, image detail, and stop behavior, and watch memory reporting and long-session compaction for fewer…
Nearby episodes from Ollama
- LFM3 Speech, Video and Reliability Fixes
- Scheduler Memory Fixes and Faster Drafting
- Mac Prefill Gains and Scheduler Fixes
- Weekly Recap - Phonon Voice Goes GA and MLX Speed Gains
- Tool Call Fixes and Phonon Voice Update
- Phonon Voice Goes Stable and API Failures Get Fixed
- Silent Failure Fixes and Mac Expansion
- API Compatibility and Runtime Reliability