Ollama 0.40.2 MLX preview
- File
- Ollama-0.40.2-40-gdb2f93e-mlx-preview-v0.40.2.dmg
- Size
- 202.5 MB
- SHA-256
1bc3a91853ba2e9d54f1dd1ee3948d778f2d756ceeb24139cf128f369162f49f- Upstream
- Ollama v0.40.2 @
b061384d90ff - Source
- integration v0.40.2 at db2f93ec0d47
- Highlights
-
Official Ollama v0.40.2New since v0.40.1 preview: model listings select the preferred migrated model and hide downgrade guard tags, avoiding duplicate entries. Also includes Claude launch context and autocompaction fixes.
Desktop options in this buildThe app remembers Chat or Apps as your home and whether the sidebar was open. Command-Q can hide the window while Ollama stays in the menu bar. It also includes a dedicated MLX-preview update check; switching to official Ollama remains a separate manual choice.
Included MLX preview capabilitiesNative GPT-OSS runtime with validated thinking and tool calling; Apertus 1.0 text and Apertus 1.1 Mini runtimes (see the model caveats)
Gemma 4 and Apertus 1.5 Omni image/audio with MP3, WebP, cancellation, and bounded media processing
Responses namespace tools, custom apply_patch compatibility, sparse tokenizer allocation, and desktop audio attachments.
Known limitations: Codex command fidelity, ordered-image answers on Gemma E4B, and Apertus 1.5 transcription remain imperfect. High image-memory use on Apertus 1.5 is under investigation (issue #36); the duplicated Command-Q setting is tracked separately (issue #37). See the model caveats.