Ollama released v0.40.7 on October 18, 2026. The release adds a new decision model and a set of reliability fixes for the OpenAI-compatible Responses path. Read the release notes.
New model
pplx-decider is a 27B-parameter decision model from Perplexity, fine-tuned from Qwen3.8-27B, joining Ollama's growing decision-model lineup.
Responses fixes
- Continuing an OpenAI-compatible Responses conversation no longer fails when switching between models with and without thinking, or when the history contains another provider's compaction items.
- Streaming Responses requests with client tool calls (the notes cite Codex with web search) no longer emit before open reasoning items, which previously produced misordered or duplicate tool-call events.
- Requests routed to remote models are now cancelled on ollama.com when the client disconnects, instead of continuing to run.
Other changes
ollama shownow reports which renderer and parser a model uses and where its chat template comes from; the API exposes the same values in newrenderer,parserandtemplate_sourcefields.- More draft models work with speculative decoding: llama-server infers the draft type automatically instead of requiring Ollama to identify it.
ollama launch claudeclampsCLAUDE_CODE_AUTO_COMPACT_WINDOWto Claude Code's supported range, avoiding a "window too large" warning.- llama.cpp and MLX were updated.
This report covers the October 18 release notes; earlier versions are covered in our reports on v0.40.6 and v0.40.5.