News brief · About 1 min read
Ollama 0.40.7 targets Responses compatibility and remote cancellation
The patch repairs conversation and streaming behavior, exposes runtime diagnostics and updates local inference components.
- Published here
- Source published
- Source checked
Ollama's 0.40.7 release notes list integration fixes and model support. The release header dates its first publication to October 18; later commit and asset timestamps are not a new launch date.
Listen to the news
Read this article aloud or keep listening to the news.
Preparing the reader…
Device voice
Speech is handled by your browser or device. AISubHub does not send the text to an audio API. Voice availability and playback with a locked screen vary by device.
Resuming may repeat a short phrase.
At a glance
- The notes list Perplexity's 27B pplx-decider, described as fine-tuned from Qwen3.8-27B.
- Responses conversations can continue when switching thinking modes or carrying another provider's compaction items; streaming fixes address misordered or duplicate client-tool events.
- Remote requests are cancelled on ollama.com when the client disconnects.
- ollama show and the API expose renderer, parser and template_source information; llama-server can infer more speculative draft-model types automatically.
- ollama launch claude constrains CLAUDE_CODE_AUTO_COMPACT_WINDOW to the supported range. The release also updates llama.cpp and MLX.
Why it matters
Our reading is that the practical focus is keeping client histories, tool events and remote work aligned. Cancellation of a remote request should not be confused with an entirely local execution path or a billing guarantee.
Limits and availability
These are project release claims, not our runtime measurements. The notes do not quantify performance, establish every model's account or hardware requirements, or guarantee compatibility for all clients.
Sources and evidence
Source published: · Source checked:
AI-assisted editorial articles checked against linked primary sources. Vendor claims are attributed; inclusion is not endorsement.