AI Agent News/stories/recursive-releases-flint-an-open-weight-40b-moe-local-coding-model-for-agents.md

Infrastructure

Recursive releases Flint, an open-weight 40B MoE local coding model for agents

Recursive has released Flint, an Apache 2.0 licensed local coding model with 40 billion total parameters and 2.5 billion active parameters. The company reports a 74.2% score on SWE-bench Verified and more than 125 tokens per second on an M4 Pro MacBook Pro.

Source: www.marktechpost.com/2026/10/21/recursive-flint-local-coding-model

Recursive has released Flint, a coding model available under the Apache 2.0 license. The model has 40 billion total parameters, of which 2.5 billion are active, and it uses a mixture-of-experts (MoE) design.

Recursive reports a score of 74.2% on SWE-bench Verified, a benchmark for resolving real software issues. It also reports more than 125 tokens per second on an M4 Pro MacBook Pro, a speed figure for running the model on a laptop.

Why it matters if you build agents

Developers building coding agents now have another open-weight model that can run locally under a permissive license. Local inference could reduce dependence on hosted model APIs for some agent workloads.

The benchmark and speed figures are Recursive's own reported numbers, so teams should test Flint on their own tasks and hardware before relying on them.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe