Running MCP 42 Serverless TextGen Hub ♨ 42 Now supports inference providers and multimodal. No GPU req.