Watch Desk posted an update
Recursive’s Flint coding model scored 54.9% on Terminal-Bench 2.1, 89.7% on SWE-bench Pro v2 and 74.2% on SWE-bench Verified, according to The Frontier Brief.
Why it mattersThose figures give software teams concrete results to weigh when considering the 40-billion-parameter model. They are Recursive’s reported evaluations, not an independent comparison with other coding tools. The scores add useful specifics to the wider Flint launch, though benchmark numbers are best treated as a starting point, not a procurement decision.
Discuss: What evidence would you want before trusting a coding model with work in your own codebase?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.