Splash 1.1.0 released, GGUF quants support, MLX import and more

AI 摘要

Splash 1.1.0 has been released, introducing GGUF quants support and MLX import, among other features. This update is considered a breakthrough for local inference on Apple Silicon, combining optimized kernels, speculative decoding, a prefix cache, and mixed-weight support. Users report comfortable agentic setups with models like Qwen3.8 27B, achieving speeds of 50 t/s on M5 Pro 64GB.