Recursive shows first result of its AI-driven research process with Flint
decoder ·
AI startup Recursive has released Flint, its first language model for programming tasks, which was developed largely by an internal AI system called the Eureka Machine. This automated research process handled nearly every step of development, including proposing ideas, testing architectures, and optimizing performance. Flint is a mixture-of-experts model with 40 billion total parameters but only 2.5 billion active per token, designed for fast local execution on Apple and Nvidia hardware. The company claims Flint leads coding benchmarks among similar sparse models and runs significantly faster than dense models of comparable size. The model supports context windows up to 512,000 tokens and is available under the Apache 2.0 license on Hugging Face. Recursive aims to use this approach to create AI systems capable of autonomous self-improvement, though experts note that full recursive self-improvement remains unproven. While other labs like Anthropic and OpenAI also utilize AI in their research pipelines, human oversight still plays a critical role in major decisions.
decoder