GoKawiilOllama update adds Ripple Preview coding model and Qwen3 memory cuts
Ollama, the open-source platform for running large language models locally on macOS, Linux and Windows, has released an update adding support for a new 65B 'Ripple Preview' model aimed at agentic coding tasks. The release also cuts memory usage for Qwen3.8 Flash Next by about 27 GiB on Apple Silicon, improves Claude Code integration, and adds CUDA 13.5 support for Windows on ARM.