Agent: Claude Code, CodexLLM: Qwen3.8, Claude 3.5#local-ai#llm-inference#apple-silicon#swift#mlx
Native Swift inference engine for Apple Silicon that streams LLM experts from SSD to run massive models like Qwen3.8-Flash-Next (125B parameters) on Macs with limited RAM. Fully offline, no Python. Compatible with Ollama, OpenAI, and Anthropic APIs. Designed for local AI coding with Claude Code, Codex, and other agents.