Ollama: Decision Models Take Shape

Decision-model work dominated activity with new Clef and Strands support, capability filtering, a dedicated ask command, and test coverage. Tool enums, onboarding polish, and proxy fixes round out reliability improvements.

Duration: PT2M28S

Episode overview

This episode is a short developer briefing from Ollama.

It explains recent repository work in plain language.

  • Show: Ollama
  • Published: 2026-10-02T13:01:20Z
  • Audio duration: PT2M28S

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 Friday, October 2nd, 2026, and this is your Ollama briefing.

The big story is decision models growing from experiment into a supported workflow, with coordinated changes across serving, discovery, and tooling.

Five related efforts landed or opened this week. Merged support for Clef in P R 18741, plus a proposal for Strands Decider support in P R 18748, expands the roster of specialized decision models. To keep clients honest, P R 18737 now reports only the decision capability for those models, so show and list won't…

For developers, the takeaway is practical: don't route conversational traffic to these models, and expect more consistent discovery and benchmarking as the framework hardens. That separation reduces misuse and should make model lists and capability checks more predictable across SDKs and automation.

The second theme is correctness and first-run polish. P R 18745 fixes tool parameter enums being rendered as plain strings, so a temperature unit list like celsius or fahrenheit now reaches the model as a literal choice, which should improve tool-call reliability for Harmony-style templates. In the desktop app, one…

Outside core code, five community…

Nearby episodes from Ollama

  1. Pointer-Head Scoring and Reliability Fixes
  2. System One Moves to Explicit Capabilities
  3. System One Decisions Launch
  4. Silent Sampling Overrides Fixed
  5. Weekly Recap - Single-Pass Structured Outputs and MLX Speedups
  6. Tool-Call Reliability Fixes
  7. Desktop Fixes and API Correctness
  8. Mac Polish, MLX Reliability, and Compatibility Fixes