Onicai brings open-source decision models fully on-chain
Onicai has released version 0.20.0 of llama_cpp_canister, opening the door to running open-source decision models fully on-chain on the Internet Computer.
Unlike conventional language models that generate text which applications then need to interpret, decision models are designed to evaluate a set of options and return a structured result. Depending on the model and task, that can include a choice, a score or a yes/no decision, along with probabilities.
The approach could make AI-driven decisions easier to use in on-chain applications, where predictable and machine-readable outputs can be more useful than free-form text.
The latest release builds on Onicai’s work to run llama.cpp as a smart contract on the Internet Computer. The project allows GGUF-format language models to be deployed to canisters and inference to be performed on-chain. Its documentation describes the system as a way to create verifiable AI inference for on-chain applications.
A decision model can be useful when an application needs an answer that fits a defined structure. Rather than asking a model to explain which option it prefers and then relying on another program to interpret the response, the model can assess the supplied choices and return a result that software can process directly.
For example, an application could ask a model to select the most suitable option from several choices, assign scores to possible outcomes or determine whether a particular condition is met. Probabilities can provide additional information about the model’s confidence in its decision.
This creates potential uses across on-chain AI agents, automated workflows and applications that need AI outputs to be checked and acted on by smart contracts.
Onicai’s llama_cpp_canister is open source under the MIT licence. The project currently supports running different GGUF models on the Internet Computer, with Qwen3-0.6B listed as its recommended default model. The project also documents larger models and reproducible builds that allow users to verify the WebAssembly deployed to a canister.
The project has been evolving through regular releases, with version 0.19.0 recorded in September. Its development has also focused on areas including performance, reproducible builds, model support and testing.
The release does not mean every AI decision can now be carried out cheaply or instantly on-chain. The Internet Computer imposes instruction and memory constraints on canister execution, and the project’s own documentation notes that model size, context length and generation limits affect performance and cost.
Still, structured decision-making offers a different way to think about on-chain AI. Instead of putting a conventional chatbot inside a smart contract, developers can use models for specific decisions where the output has a defined format and can feed directly into an application.
With llama_cpp_canister continuing to bring open-source models into the Internet Computer environment, Onicai is positioning decision-focused AI as another building block for verifiable, on-chain applications.
Dear Reader,
Ledger Life is an independent platform dedicated to covering the Internet Computer (ICP) ecosystem and beyond. We focus on real stories, builder updates, project launches, and the quiet innovations that often get missed.
We’re not backed by sponsors. We rely on readers like you.
If you find value in what we publish—whether it’s deep dives into dApps, explainers on decentralised tech, or just keeping track of what’s moving in Web3—please consider making a donation. It helps us cover costs, stay consistent, and remain truly independent.
Your support goes a long way.
🧠 ICP Principal: ins6i-d53ug-zxmgh-qvum3-r3pvl-ufcvu-bdyon-ovzdy-d26k3-lgq2v-3qe
🧾 ICP Address: f8deb966878f8b83204b251d5d799e0345ea72b8e62e8cf9da8d8830e1b3b05f
Every contribution helps keep the lights on, the stories flowing, and the crypto clutter out.
Thank you for reading, sharing, and being part of this experiment in decentralised media.
—Team Ledger Life