Goldie and Mirhoseini's chip-design lab released a 65B dense model it says scored 80.5 on TerminalBench 2.1 and cost $32M to train.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
· Updated
Primary source: [x.com](https://x.com/RicursiveAI/status/2118046765857546733)
Why it matters #
Ripple is the first public model from a $4B chip-design startup built by AlphaChip's creators, and its own model card both supports and limits the efficiency claim. If independent runs confirm the TerminalBench result at 65B, developers get a locally deployable open-weight coding model that challenges much larger systems on one key agentic test.
Ricursive Intelligence released Ripple-65B Preview on November 4th, a 65-billion-parameter dense language model that co-founders Anna Goldie and Azalia Mirhoseini built to power the Palo Alto company's AI chip design platform. They are now offering it as a general-purpose open-weight model. Ricursive announced the release in a thread on X and on its Ripple page, saying Ripple is a general-purpose variant of the proprietary model behind its chip design work.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.