Ollama: Rendering Transparency and Response Fidelity Fixes
This week centered on making model packaging observable and stopping silent output loss. Show and create now expose rendering decisions, while a cluster of Anthropic, OpenAI, and parser fixes restores dropped thinking, tool calls, and timestamps.
Duration: PT2M33S
Episode overview
This episode is a short developer briefing from Ollama.
It explains recent repository work in plain language.
- Show: Ollama
- Published: 2026-10-18T13:01:03Z
- Audio duration: PT2M33S
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 18th, 2026.
The throughline this week is visibility and fidelity — Ollama is revealing more about how models are built, while fixing cases where valid model output never reached you.
First, packaging transparency. The show command now actually fills in renderer, parser, and a new template source field, so you can see what renders chat prompts — a configured renderer and parser, a Go template layer, embedded Jinja in the model file, Harmony, or none. That's PR 19095. Paired with that, creation in…
Second, response correctness across provider A-P-Is and parsers. Two Anthropic fixes cover web search reusing the relaxed thinking setting from PR 19098, and streaming thinking that arrives after text or tool use in PR 19110, which was previously dropped for streaming clients. For OpenAI compatibility, PR 19106…
Finally, core reliability. PR 19096 restarts downloads cleanly when servers ignore byte ranges, avoiding repeated chunks and digest errors. PR 19097 stops reuse of runners cancelled during startup, and PR 19091 requires the embedding capability up front for embed endpoints with a clearer error.
What's next: if you build or debug models, check show output…
Nearby episodes from Ollama
- Remote Cancel Fixes and Wider Model Support
- MLX Compatibility Push and Responses Fixes
- Weekly Recap - Decision Models Go Live, MLX Gains, and API Polish
- Proxy Transfers and API Resilience
- Model Compatibility Catch-Up and Windows Reliability
- Responses Compatibility and Clearer Errors
- Decision Models Go Usable, MLX Speeds Up on CUDA
- Decision Scoring Matures as Inference Hardens