what stackpulse tracks
Ollama releases from GitHub
StackPulse watches Ollama release notes and keeps the original source link close to every summary.
#stacks/
loading releases…
releases
breaking
security
deprecated
migrations
Get up and running with large language models locally StackPulse turns upstream changelogs into scannable summaries with risky changes, deprecations, migration notes, and source links.
what stackpulse tracks
StackPulse watches Ollama release notes and keeps the original source link close to every summary.
upgrade risk
Risky changes are separated from normal feature notes so you can scan upgrade impact before changing production dependencies.
migration notes
Migration steps and recommended actions are only shown when the upstream release notes support them.
This release introduces automatic MLX runtime usage for supported models on Apple Silicon devices, potentially improving performance. Additional models will be enabled during the pre-release testing phase.
Apple Silicon users will see improved performance as supported models now default to the MLX runtime.
Apple Silicon users should test their models with the new MLX runtime default.
This release focuses on performance improvements and bug fixes, particularly for Apple Silicon devices. Key changes include faster structured outputs for thinking models, optimizations for Qwen 3.8 prompt processing, and enhanced image resolution handling for Gemma 4.
Users on Apple Silicon devices and those using structured outputs or high-resolution image processing will benefit most from these changes.
This release primarily focuses on bug fixes and performance improvements. It resolves intermittent 'model not found' errors and optimizes structured output processing for thinking models.
Users experiencing 'model not found' errors or using thinking models with structured outputs may be affected.
Update to this version if experiencing 'model not found' errors or for improved performance with certain models.
This release introduces enhanced thinking controls for models, adds support for Nemotron H vision models on Apple Silicon, and includes fixes for model pulls from HuggingFace.
Users leveraging model thinking controls or running Nemotron H vision models on Apple Silicon are directly affected.
Update to v0.34.3 to benefit from new features and fixes.
This release introduces support for Nemotron H vision models on Apple Silicon with MLX, improves the macOS app behavior, and fixes model pulls from HuggingFace.
Users leveraging Apple Silicon, macOS app, or HuggingFace model pulls are affected.
This release introduces new features including thinking controls in the API, support for Nemotron H vision models on Apple Silicon, and improvements to the macOS app behavior.
Users of the Ollama API and macOS app are affected by the new features and improvements.
Update to v0.34.3-rc0 to take advantage of the new features and improvements.
This release primarily includes updates to the underlying llama.cpp library, with no detailed changes provided in the release notes.
view source on github->This release includes updates to llama.cpp, but no specific details are provided in the release notes.
view source on github->This release introduces a first-run setup for `ollama`, adds direct app page access on macOS and Windows, and fixes memory growth issues during long generations.
Users running `ollama` for the first time or on macOS/Windows are primarily affected.
Update to v0.34.2 to benefit from the new setup flow and memory fixes.
This release focuses on improving MLX memory handling on Apple Silicon, optimizing model creation, and enhancing API performance. It also introduces stricter repeat token detection and deprecates `typical_p` for new models.
Users creating GGUF models or relying on `typical_p` for new models are affected.
Update GGUF model creation workflows to use llama.cpp tooling and avoid setting `typical_p` for new models.
This release focuses on memory management improvements for MLX models, increases token limits, and includes UI/UX enhancements. It also updates underlying MLX and llama.cpp dependencies.
Users of Ollama's MLX and llama.cpp integrations may experience improved memory management and model loading behavior.
This release focuses on improving memory management, model loading, and user interface fixes. Key changes include evicting prefix cache snapshots, checking system free memory before loading models, and raising the token repeat limit.
Users relying on MLX models or experiencing memory constraints during model loading will be affected.
Update to this release for improved memory management and model loading stability.
This release introduces integration with ChatGPT Desktop for direct use of Ollama models, improves performance on Apple Silicon, and adds support for OpenAI-compatible client tools and image handling.
Users of Ollama models and ChatGPT Desktop, especially those on Apple Silicon, are affected by these changes.
Update to v0.34.0-rc3 to take advantage of the new features and improvements.
This release introduces integration with ChatGPT Desktop, allowing Ollama models to be used directly within the ChatGPT workflow. It also includes performance improvements for Apple Silicon and enhancements for OpenAI-compatible client tools.
Users of ChatGPT Desktop and Apple Silicon devices will benefit from the new features and performance improvements.
Update to the latest version to take advantage of the new integrations and performance enhancements.
This release adds integration with ChatGPT Desktop, allowing users to run Ollama models directly within the ChatGPT interface. It also includes performance improvements for Apple Silicon and adds support for OpenAI-compatible client tools.
Users of Ollama and ChatGPT Desktop on MacOS are affected by the new integration feature.
Update to v0.34.0 to take advantage of the new ChatGPT Desktop integration and performance improvements.
This release primarily updates support for GGUF models by honoring their default parameters and includes updates for MLX, MLX-C, and llama.cpp integrations.
Users of GGUF models and those utilizing MLX or llama.cpp integrations may be affected by the updates.
This release adds support for images and audio in the Gemma4 model on the MLX engine, improves token reporting, and updates several dependencies including MLX and llama.cpp.
Users of the Gemma4 model or those utilizing MLX engine features may benefit from the updates.
This release focuses on bug fixes and improvements, including restoring system dark mode, handling model catalog changes gracefully, and synchronizing macOS app handoff.
Users relying on dark mode, macOS app handoff, or model catalog changes are affected.
Update to this version to benefit from the fixes and improvements.
This release focuses on improving user experience by restoring dark mode support, fixing macOS app behavior, and enhancing the Claude Desktop proxy.
Users of the macOS app and those relying on dark mode or the Claude Desktop proxy are affected.
Update to the latest version to benefit from the improvements.
This release introduces Qwen3.8 Flash Next support for MLX, adds structured output capabilities to mlxrunner, and improves Metal GPU timeout handling when loading models from slow storage.
Users leveraging MLX or llama.cpp integrations may benefit from performance improvements and new features.