Ricursive launches Ripple-65B Preview, a general‑purpose AI model for advanced chip design under Apache 2.0, priced at $0.60 per million outputs (Carl Franzen/VentureBeat)
chips
| Source: Techmeme | Original article
Ricursive has launched Ripple-65B Preview, an open‑source, general‑purpose AI model for automating advanced chip design, available under Apache 2.0 and priced at $0.60 per million outputs.
Ricursive Intelligence, the Palo Alto lab behind the AlphaChip reinforcement‑learning system for chip floor‑planning, has opened up its latest language model, Ripple‑65B Preview. The 65‑billion‑parameter dense model is released under an Apache 2.0 licence, making the weights freely downloadable. It is a general‑purpose variant of the proprietary model that powers Ricursive’s AI‑driven chip‑design platform.
The company says Ripple‑65B Preview was trained from scratch on 12 trillion tokens and supports context windows of up to one million tokens. In its own benchmarks the model scores 80.52 on Terminal‑Bench 2.1, a result the firm claims is the highest among U.S. open‑weight models on two demanding coding tests. Hosted inference is priced at $0.60 per million output tokens, positioning the model as a low‑cost option for developers who need large‑scale code generation or other agentic tasks.
The release matters because it blends the efficiency gains of a chip‑design‑focused architecture with a permissive licence, potentially lowering the barrier for research and commercial teams to build advanced coding assistants without relying on closed‑source offerings. By publishing the weights, Ricursive also invites community contributions that could accelerate optimisation for a broader set of applications beyond hardware design.
Going forward, observers will watch how Ripple‑65B Preview performs on real‑world software development workloads and whether third‑party services adopt its pricing model. Equally important will be any follow‑up releases that extend the model’s capabilities or integrate it with Ricursive’s existing chip‑design tools, signalling how the company plans to leverage the open model to reinforce its niche in AI‑assisted hardware engineering.
Sources
Back to AIPULSEN