Typical answer
Context, caveats and a recap around a few lines of code.
Nock Coder is a small open model that answers with code first and one short sentence after. It runs on your own machine, with no account and no API key.
Early preview · Built on Qwen2.5 Coder · Apache 2.0
Ask for a snippet, a fix or a contract. Nock Coder answers with the code and stops, so you spend less time scrolling and more time shipping.
You asked for ten lines of code and got three paragraphs of context, a list of caveats and a recap. We trained Nock Coder to skip all of it.
Typical answer
Context, caveats and a recap around a few lines of code.
Nock Coder
def reverse(text: str) -> str:
return text[::-1]
The code, then one sentence.
Illustration of the answer style, not a measurement.
The 1.5B model compresses to a GGUF file of roughly one gigabyte. Run it with Ollama or llama.cpp on Windows, macOS or Linux. Your code never leaves your computer.
# 1. Download from Hugging Face
huggingface-cli download Nock-AI/nock-coder-1.5b-GGUF --local-dir nock-coder
# 2. Create the model in Ollama
cd nock-coder && ollama create nock-coder -f Modelfile
# 3. Ask it something
ollama run nock-coder
Short answers. Your own machine. No account to create and nothing to send anywhere.
Each model starts from the base that fits its task and gets its own name. We ship them one at a time.
Writes and fixes code in many languages, including Solidity for Robinhood Chain. Base: Qwen2.5 Coder 1.5B Instruct.
Translates crypto docs and wallet prompts into 8 languages and keeps terms consistent. Base: Hy MT2.
Describes what a transaction or contract does in plain language. Base to be announced.
We plan to run common coding benchmarks and a short set of Solidity tasks, then release the raw outputs next to the numbers. Until then, nothing here is a claim.
Python and general code
Pending
Solidity tasks
Pending
Answer length versus base model
Pending
Short answers about the model, running it and licensing.
NockAI is a set of small language models for people who build on and use Robinhood Chain. The first one is Nock Coder, a coding model. More follow once it is tested.
Nock Coder is a LoRA fine tune of Qwen2.5 Coder 1.5B Instruct, which is licensed under Apache 2.0. We trained it on short, code first answers.
No. NockAI is an independent project. Robinhood and Robinhood Chain are trademarks of their owners and are named only to describe what the models are for.
Yes. We publish a quantized GGUF build for llama.cpp and Ollama. At 1.5 billion parameters it runs on a typical laptop, including one without a dedicated graphics card.
It is a small model, so expect mistakes on hard problems. We have not published benchmark scores yet, and the evaluation section shows what we will measure. Review any code before you run it.
No model output should go on chain without review. Nock Coder is a drafting aid. Have a qualified person audit any contract before you deploy it.
Yes. We release the weights openly and state the license on each model card.
Short code first answers drawn from two MIT licensed instruction datasets, one general and one for Solidity, plus a few identity examples. Both datasets contain text generated by other language models, and the model card lists them.
The weights, the GGUF build and the model card live on Hugging Face. Download it and try it on your own machine.