← EVENT LIBRARY / AI Engineering

Ollama adds probability-first decision models for local routing

Ollama now serves local decision models, streaming speech paths and Liquid AI LFM3 models. Version 0.40.13 expands LFM3, including lfm3-audio text-to-speech and lfm3-omni, from Apple silicon to Linux and Windows. Model quality, hardware-specific latency and individual model licenses still require workload review.

Availability: local Ollama runtime

What this means for you

For teams doing local classification or routing, structured probabilities are easier to calibrate, threshold and fall back than labels parsed from generated text. Test calibration error, abstention thresholds and latency on labeled data before relying on confidence values.

Evidence & related material

Supporting evidence

Ollama 0.35.0 stable adds `/v1/systemone`, letting decision models return choices, probabilities and scores instead of generated text. Nimble and Tev1 are the first supported models for ticket triage, model routing and classification. The API exposes choice, noul and score questions. The prior run mistakenly treated this stable release as prerelease, so this run records it as a late discovery.

Subsequent changes

Update · 2026-09-30

Ollama 0.35.1 lets clients recognize decision-model capability

Update · 2026-10-22

Ollama 0.40.9 adds local streaming speech on Apple Silicon

Update · 2026-10-29

Ollama 0.40.13 expands LFM3 audio and omni models to Linux and Windows.

Technical details & scores
Version 0.35.0 stable
Endpoint /v1/systemone
Outputs choices, probabilities, confidence and ordered scores
Models Nimble and Tev1
Question Types choice, noul and score
Availability local Ollama runtime
Latest Version 0.40.13
Platform Update LFM3, lfm3-audio and lfm3-omni now run on Linux and Windows as well as Apple silicon
Time Basis GitHub stable release published 2026-10-28T19:55:04Z
Technical impact 3
Engineering value 4
Adoption signal 3
Maturity 4
Verification cost 2
Risk 2