Ricursive releases Ripple-65B Preview, an Apache-licensed open-weight model built for agentic coding
Ricursive Intelligence has released Ripple-65B Preview, a 65-billion-parameter dense open-weight model pretrained for agentic coding. The company claims a top self-reported Terminal-Bench 2.1 score among US open-weight models, though the figures have not been independently reproduced and some reruns narrow the gap.
Ricursive Intelligence released Ripple-65B Preview on 4 November, a 65-billion-parameter dense model whose weights are published under the Apache 2.0 licence. The company describes it as a general-purpose variant of the proprietary model behind its chip-design platform, pretrained from scratch with agentic coding in mind.
The headline claim is a score of 80.52 on Terminal-Bench 2.1, which Ricursive presents as the highest self-reported result among US open-weight models. It also says the model cost roughly $32 million to develop and train, about a tenth of the previous leader in the category. Both figures come from the company itself.
The weights, a download of around 130 GB, are available on Hugging Face, Together AI, GMI Cloud and OpenRouter. The model card advertises context of up to one million tokens, although at least one hosted endpoint currently serves far less.
Caveats matter here. The qualifier "US" excludes labs such as DeepSeek and Qwen. When Ricursive reran the benchmark with the official harness at 128K context, a rival's score rose above 78 while another fell to roughly 71, and on DeepSWE 1.1 one analysis puts Ripple well behind Meta's Muse Wonder. Treat the numbers as provisional until independent reruns land.
Why it matters
A dense 65B model with permissive licensing is small enough to host on a single high-end workstation or modest GPU cluster, which changes the economics of running coding agents privately. If the training-cost claim holds, it also suggests frontier-adjacent capability is getting cheaper to reproduce.
Indicops view
For Indian enterprises and public bodies with data-residency constraints, capable open-weight coding models are worth piloting on in-country infrastructure, but only after internal evaluation on your own repositories rather than on vendor benchmarks.
More in Open Source
Back to the newsroomHugging Face's chief vows to fight on for open source as Nvidia's $12.9 billion bid faces uncertainty
After Jensen Huang's exit as Nvidia CEO, Hugging Face chief Clement Delangue said the company will keep fighting for open-source AI harder than ever. Nvidia's planned $12.9 billion purchase of the platform is scheduled to close in the first half of 2027 and has drawn US Justice Department questions.
Nvidia donates OpenShell agent runtime to CNCF and previews GPU Kubernetes tooling
Nvidia has donated OpenShell, an open-source runtime that applies policy controls to autonomous AI agents, to the Cloud Native Computing Foundation. At KubeCon North America it also put new GPU Kubernetes tools and a fast inference-restore feature into public preview.
Baidu releases ERNIE 5.5 Flash Base, a 280-billion-parameter open-weight model with a restrictive licence
Baidu has published ERNIE 5.5 Flash Base, a 280B-parameter mixture-of-experts checkpoint with 14B active per token, trained on 33.5 trillion tokens. Weights are downloadable, but the licence excludes the EU, UK and South Korea and limits commercial use above $50 million revenue.