Recursive published Apache-2.0 Flint weights and Ollama 0.40.8 provides a macOS package; performance remains publisher-reported.
Recursive releases Flint as an open-weight sparse coding model
2026-10-22Frontier watchTRIAL
Recursive released Flint, a 39.8B-parameter mixture-of-experts coding model with about 2.5B active parameters per token and a 262K context window. BF16 weights are Apache-2.0, with SGLang and vLLM instructions; Ollama 0.40.8 adds a roughly 23 GB package for macOS. Other Ollama GPU support is promised later, and benchmarks remain publisher-run.
What this means for you
For teams running coding models locally, Flint offers a small active footprint with a long context and permissive weights. A bounded trial should measure patch validity, latency, memory and repeated-run variance.
FRONTIER WATCH
Public artifacts and validated results are different.
The recorded fact is the public work and its artifacts. Author-reported outcomes do not establish independent reproduction. Next checks include usable code, task conditions and third-party validation.
Evidence & related material
Technical details & scores
| Parameters | 39.8B total, about 2.5B active per token |
|---|---|
| Context | 262K tokens |
| License | Apache-2.0 weights |
| Serving | SGLang and vLLM instructions; about 23 GB Ollama package on macOS |
| Hardware | publisher says one H200 or B300 for BF16; other Ollama GPU support follows later |
| Reported Benchmarks | 54.9 Terminal-Bench 2.1, 89.7 SWE-bench Pro v2, 74.2 SWE-bench Verified; publisher-run and harness-dependent |
| Time Basis | Ollama v0.40.8 stable published 2026-10-21T16:20:30Z |
Technical impact 3
Engineering value 4
Adoption signal 2
Maturity 3
Verification cost 3
Risk 2