The company says the model has 40 billion total parameters and activates 2.5 billion per token. Its weights are available on Hugging Face under the Apache 2.0 license.
Recursive reports more than 125 tokens per second on an M4 Pro MacBook Pro and more than 150 tokens per second on a 64GB NVIDIA RTX Spark at short contexts, using its optimized inference engine. Flint supports up to a 512k-token context window; Recursive recommends staying within 256k tokens for coding and other tasks.
The release names llama.cpp, Ollama, Unsloth, LM Studio, vLLM, and SGLang as compatible tools. Recursive also released rec, a lightweight inference engine optimized for Flint on M4 Pro and M5 Pro/Max hardware. The company says its Eureka Machine contributed to the model's data, training recipes, infrastructure, and inference optimization, and that a small team built Flint in under four months.
Recursive reports benchmark leadership among sub-50B MoE models and comparisons with dense 25B-40B models. The source does not establish those results through an independent evaluation, nor does it specify performance or hardware requirements for every local deployment.