By No Code MBA · Facts checked
- Provider
- Recursive
- Announcement date
- Status at last check
- Generally available
- Model family
- Flint
- Capabilities
- Text, Coding
What changed
Flint is a mixture-of-experts language model for coding agents that Recursive says was trained from scratch. It has 39.8B total parameters with about 2.5B active per token, 256 routed experts per layer (6 active plus a shared expert), and a documented 262,144-token context covering prompt and output. Recursive reports 54.9% on Terminal-Bench 2.1, 74.2% on SWE-bench Verified and 89.7% on SWE-bench Pro v2. It is text in, text out, with reasoning and multi-turn tool use.
Access and availability
Weights are on Hugging Face at RecursiveSI/flint under Apache 2.0. The BF16 checkpoint is about 80 GB and Recursive's included SGLang kit serves it on one B300 or H200 GPU per replica; no Hugging Face inference provider hosts it.
Official access documentationWhat you can build
Evaluate it on a small, representative repository task, such as diagnosing a failing test and proposing a fix, then verify every change with your usual tests and review process.
Limitations
Recursive warns it can produce faulty code or inappropriate tool calls, so code should be tested and tool-using apps need their own permission boundaries. Long reasoning adds latency and can exhaust the output budget, and Transformers modeling code is not included yet.