5 November 2026
Ricursive previews 65B open model, claims top US coding score
First reported
Latent Space ran this on .
- Agentic coding means an AI completing multi-step programming tasks on its own, and the claim rests on the Terminal-Bench 2.1 test.
- The company says the model narrowly beat Beam-501B, and that it cost a stated $32 million to build.
Our read
Only Latent Space picked this up among the 30 newsletters tracked. It follows the Argon coding and robot results we covered on 2026-10-14, where reporting questioned benchmark claims.
Earlier on this story
Written by Newskeryx from which newsletters covered this story and from our own archive.
How it was covered
Latent Spaceswyx & Alessio
Ripple-65B benchmarks, training budget and scaling ablations, and critics' objections