Recursive/Released Oct 18, 2026

Flint

Not ranked · too few of its benchmarks are in the capability model to estimate a score

Flint is Recursive's open-weight mixture-of-experts language model for coding agents, trained from scratch, with 39.8 billion total parameters and about 2.5 billion active per token, counting the output head. Its 49 layers (2 dense, then 47 that route each token to 6 of 256 experts plus a shared expert) mix 35 sliding-window layers with 14 full-attention layers. It takes and returns text, decides on each turn whether to reason, with no switch or budget, supports tool calls, and is configured for 262,144 tokens of context. The launch post is dated October 18, 2026, although its Hugging Face repositories were created on October 21, 2026.

Scored on 2 benchmarks · 18 total scores · 18 independently reported