Recursive 使用 Eureka Machine 構建 Flint，一款 400 億參數的 AI 模型

recursive, an ai startup, has released flint, its first language model developed largely by an internal ai system called 'eureka machine'. flint uses a mixture-of-experts architecture with 40 billion parameters, 2.5 billion active, and is optimized for fast local execution on apple and nvidia hardware. the model was built in four months and leads coding benchmarks like swe-bench pro v2 and terminal-bench 2.1 among moe models with fewer than 50 billion parameters. recursive claims flint matches benchmark scores of much slower dense models in the 27 to 31 billion parameter range and generates over 125 tokens per second on an m4 pro.