ollama releases

Details

Performance Improvements
- Structured outputs in reasoning (thinking) models are now applied in a single pass, improving processing speed and reliability.
- Prompt processing speed for Qwen 3.8 has increased in Apple Silicon environments.
- Gemma 4 on Apple Silicon now selects the optimal resolution per image, better preserving details in high-resolution images.
Bug Fixes
- Fixed intermittent "model not found" errors that occurred when the local model library was large.
- Resolved an issue where the macOS app became unresponsive when checking if ChatGPT or Codex was running.
Dependency Updates
- Updated the llama.cpp, MLX, and XGrammar libraries.

Performance Improvements
- Structured outputs in reasoning (thinking) models are now applied in a single pass, improving processing speed and reliability.
- Prompt processing speed for Qwen 3.8 has increased in Apple Silicon environments.
- Gemma 4 on Apple Silicon now selects the optimal resolution per image, better preserving details in high-resolution images.
Bug Fixes
- Fixed intermittent "model not found" errors that occurred when the local model library was large.
- Resolved an issue where the macOS app became unresponsive when checking if ChatGPT or Codex was running.
Dependency Updates
- Updated the llama.cpp, MLX, and XGrammar libraries.

ollama