GitHub radar
Ripple-65B: a local AI coding assistant for your own machine
It's a 65-billion-parameter open-weight language model you can download and run on your own hardware, with no cloud involved. It writes, reads, and debugs code, even in very long files.
A large AI model you run locally — it reads, writes, and fixes code, even in huge files.
When it helps
Useful if you want a powerful coding assistant without the cloud, on a strong Mac or a PC with a good graphics card. Skip it on a regular laptop or for small one-off tasks.
Pros
- Runs fully locally, no internet or subscription needed
- Reads almost a million tokens at once — entire codebases
- Beats models 8–15x larger on agentic coding tests
- Free Apache 2.0 license
Cons
- Needs a Mac with 64GB+ memory or a 48GB GPU
- It's an early preview, not polished for casual chat
- Benchmark numbers come from the maker, not independent tests
How to set it up — step by step
- 1Check your hardware: you need a Mac with 64GB+ memory or a 48GB GPU, or it won't run
- 2Open your AI agent (Claude Code, Codex) inside your project folder
- 3Send the agent the text from the block below
- 4Confirm the agent found a compressed model version and launched it via Ollama or llama.cpp
- 5Ask it to write and debug a small test program to prove the model works
- 6Keep in mind the benchmark scores are self-reported — verify results yourself
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the Ripple-65B Preview model by Ricursive Intelligence for me: https://huggingface.co/ricursiveAI/Ripple-65B-Preview — find a ready quantized version (like NVFP4) that fits my hardware, run it through Ollama or llama.cpp, and show me how to ask it to write and debug a small program.
You'll need a Mac with 64 GB of memory or more, or a GPU with 48 GB of VRAM — a regular laptop won't run this one.
Open on Hugging Face▌ More finds