Breaking02 / IT
Ollama v0.40.11 brings macOS speech generation and d3 models
The release notes state no licence and give no VRAM requirements for each d3 size

Ollama released v0.40.11 on 24 Oct 2026 and said its speech generation runs on macOS, in release notes that do not state a licence.
Speech. According to the release notes, ollama run plays audio as it is generated. It reads aloud text or a model's reply.
The notes give the OpenAI-compatible /v1/audio/speech endpoint as the streaming route for apps. The example command there uses the phonon model.
d3. Ollama said d3 is a family of decision models from vLLM Semantic Router. Inputs are text, structured state or images. Outputs are classifications, yes-or-no answers and probability scores for options.
The release notes list d3 in 2B, 4B, 9B and 27B sizes. The full changelog links to a comparison with v0.40.10.
The release notes do not give VRAM requirements for each d3 size or a schedule for speech support beyond macOS.