Recursive releases Flint, a 40B open-weight coding model that runs locally
Recursive Superintelligence published Flint on October 21, an open-weight coding model under the Apache 2.0 license. It is a mixture-of-experts transformer with 39.8 billion total parameters and about 2.5 billion active per token, with 262,144 documented context tokens. The company claims more than 125 tokens per second on a MacBook Pro with an M4 Pro chip, though the GGUF variant for Apple Silicon requires at least 48 GB of unified memory. Recursive raised more than US$650 million and signed a US$410 million compute deal with AWS.
In 50 seconds
- Flint is a MoE with 39.8 billion total parameters and about 2.5 billion active per token, under the Apache 2.0 license.
- The company claims more than 125 tokens per second on a MacBook Pro with M4 Pro and more than 150 on a 64 GB RTX Spark in short contexts.
- The GGUF variant for Apple Silicon takes about 26 GB and requires at least 48 GB of unified memory; the BF16 checkpoint is around 80 GB.
Read the full story at El Ecosistema Startup
IASummary generated by artificial intelligence from a story by El Ecosistema Startup. It may contain errors: the original source prevails.
Comments
Be respectful. Comments reported by several people are hidden until a moderator reviews them.
You do not need an account to read, only to comment. Sign in with a code sent to your email or with your Google account. We only show the name you choose, never your email.