SYLIN

Ripple 65B Preview

by Ricursive AI · Model · Language & code models

unproven

Open 65B model claiming top US open-weight agentic coding; no independent tests yet

65B dense model built for agentic coding, with up to 1M tokens of context.

AccessOpen weightsPrice$0.1 in · $0.3 out per M tokensContext1,049K tokensNeeds~57 GB; smallest build ~39 GBDownload57 GB in FP4, 130 GB in BF16; Ollama build is 39 GBSize65B parametersLicenseApache-2.0

Why we say so

Every claim comes from Ricursive itself, including the state-of-the-art coding claim, and no benchmark or tester backs it. It is labelled a preview and was built for AI chip design. The Apache 2.0 license and downloads are real and easy to try.

Evidence

  1. Nov 14
    Dense 65B preview claims US open-weight state of the art on agentic coding
  2. Nov 8
    Dense 65B open model with 1M context, Apache 2.0; BF16 130 GB, FP4 57 GB; built for AI chip design, $32M R&D
  3. Nov 8
    Claims new state of the art for agentic coding among open models from the US

Cost to run