Recursive releases 40B open-weight model for offline local coding

Recursive released Flint, its first open-weight coding model, aimed at developers who want to work offline. The 40-billion-parameter mixture-of-experts model activates 2.5 billion parameters per token. It is designed to run locally, keep code on-device, and avoid cloud API fees. Recursive says the model can generate more than 125 tokens per second on an M4 Pro MacBook Pro and performs strongly against other smaller coding models on Recursive's own benchmarks. Flint was built in under four months with substantial help from Recursive's human-supervised Eureka Machine and is available under the Apache 2.0 license. The company cautions that the model may need more explicit guidance than larger commercial coding assistants. Recursive presents the release as an early, limited example of AI helping develop AI.