v0.30.0
Breaking Changes📦 ollamaView on GitHub →
⚠ 2 breaking✨ 5 features🔧 1 symbols
Summary
Ollama 0.30 introduces significant performance and compatibility improvements by integrating llama.cpp, broadening hardware support, and adding new model capabilities.
⚠️ Breaking Changes
- The model 'nomic-embed-text' now converts inputs to lowercase. If your application relied on preserving mixed case inputs for this specific model, you must now ensure inputs are already lowercased or handle the casing change.
- Prior Ollama versions incorrectly preserved mixed case for 'nomic-embed-text'; this behavior is now fixed by enforcing lowercase conversion.
✨ New Features
- Improved compatibility and performance via integration with llama.cpp.
- Augmented MLX engine support on Apple Silicon.
- Support for a wider range of models, including GGUF-based models from Hugging Face.
- Support for user's own fine-tuned models.
- Faster performance on NVIDIA hardware.