# Ripple: A Local AI Model Co-Designed for GPUs

- Original post: https://www.twitter.com/i/status/2118046765857546733
- Source account: [@RicursiveAI](https://x.com/RicursiveAI)
- Source language: English
- Duration: 1:14
- Published: 2026-11-05
- Canonical URL: https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus

## Transcript

[0:02](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=2) Today we're releasing Ripple, a frontier local model that outperforms US open-weight models fifteen X larger in size on agentic tasks like Terminal Bench at a size that can run locally on a MacBook Pro or DGX Spark.

[0:20](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=20) By pre-training Ripple with a deep understanding of the hardware it runs on, we were able to reach the same performance as the open source baseline with sixteen X less compute.

[0:30](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=30) It's like we trained with sixteen times more GPUs.

[0:32](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=32) We built Ripple for three reasons. First, to show what we can gain by co-designing the model with the hardware it runs on, in this case, a GPU.

[0:40](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=40) Second, to power our AI chip design platform.

[0:43](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=43) An in-house version of Ripple orchestrates and learns across our proprietary AI algorithms to quickly design more performant chips.

[0:51](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=51) Third, to start closing the loop, training better models that design better chips that train even better models.

[0:58](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=58) Ripple is our first model, and it was trained on silicon that wasn't designed for it.

[1:05](https://www.translate.mom/transcript/cFSxIKkgbOxM/en/ripple-a-local-ai-model-co-designed-for-gpus?t=65) Imagine what's possible on silicon that is.