v0.32.6-rc0Pre-release
v0.32.6
Changed 4
- Qwen3.5 is faster on Apple GPUs with the MLX engine now using the model's MTP head for speculative decoding automatically
- /v1/chat/completions streaming now matches OpenAI's wire format with role only on the first chunk, finish_reason on its own chunk, and usage in a separate chunk with stream_options.include_usage
- ollama run kimi-k3 now offers kimi-k3:cloud for cloud-only models that publish no default tag instead of failing
- Updated the MLX and llama.cpp engines
Fixed 4
- Truncated OpenAI responses now report finish_reason: "length" instead of "tool_calls"
- TUI pipe-delimited prose no longer renders as a table
- TUI Enter key now accepts the highlighted @ file completion
- TUI /prompt scrolling is no longer laggy
Removed 1
- Experimental image generation has been temporarily removed
What's Changed
- Qwen3.5 is faster on Apple GPUs: the MLX engine now uses the model's MTP head for speculative decoding automatically
/v1/chat/completionsstreaming now matches OpenAI's wire format:roleonly on the first chunk,finish_reasonon its own chunk, and usage in a separate chunk withstream_options.include_usage.- Truncated OpenAI responses now report
finish_reason: "length"instead of"tool_calls". ollama run kimi-k3now offerskimi-k3:cloudfor cloud-only models that publish no default tag, instead of failing.- TUI fixes: pipe-delimited prose no longer renders as a table, Enter accepts the highlighted
@file completion, and/promptscrolling is no longer laggy. - Experimental image generation has been temporarily removed. Continue using 0.32.5 for image generation support
- Updated the MLX and llama.cpp engines.
Full Changelog: https://github.com/ollama/ollama/compare/v0.32.5...v0.32.6-rc0