Ollama 0.32.4 adds Apple Silicon GPU support and Laguna MLX inference for top open coding model

The new Ollama release enables Apple GPU acceleration via MLX for Poolside Laguna S 2.1, currently one of the strongest open-weight coding models, along with speculative decoding improvements and Qwen3 MoE expert quantization fixes. Teams with sufficient Apple Silicon memory can now run frontier-class open coding agents locally.