Ollama – v0.21.0-rc0
Ollama just dropped a new release candidate, v0.21.0-rc0, and it’s all about fine-tuning performance for some heavy hitters! 🚀
If you’re looking to run powerful LLMs like Llama 3, DeepSeek-R1, or Gemma locally without a massive cloud budget, this is the tool to keep in your kit. This update specifically targets optimization for those of us tinkering on Apple Silicon.
What’s new in this release:
- MLX Optimization: A specific fix has been implemented for the MLX backend regarding Gemma 4 cache usage. It now utilizes a logical view, which streamlines how the model handles data during inference. 🛠️
This is a big win for memory management! If you are experimenting with the latest Gemma models on Mac, this update ensures much smoother performance and more reliable data handling during your local runs. Happy tinkering! 💻✨