Veritate, Carpathian's AI Inference Engine
Veritate is an AI inference engine that runs and trains large language models. It is tuned to work on a normal desktop CPU instead of a high-end graphics card.
Veritate is an AI inference engine that runs and trains large language models. It is tuned to work on a normal desktop CPU instead of a high-end graphics card.
Fine-tuning teaches behavior, RAG supplies knowledge, prompting shapes both. What the published comparisons found, and how to pick without burning a GPU budget.
Why AI chat can be free, what happens to what you type, and the short checklist worth running before anything from work goes into a chat box.
When owning your hardware beats renting a virtual machine, when it does not, and the questions to ask a colocation facility before you commit to anything.
Object storage explained: how it differs from a server disk, what it is good and bad at, and the signs that tell you it is time to use a bucket.
Self-hosting an open-weight model or calling a hosted API? The drivers on each side, the break-even point, and the hidden costs.
Veritate is Carpathian's Open Source AI inference and training engine. It's hand-written in C and assembly, tuned to work on a normal desktop CPU instead of a high-end graphics card.
Most modern AI tools assume you own, or rent, expensive specialized hardware. Veritate does not. It is built so the chip already sitting inside a regular consumer PC is enough.
We are building a complete pipeline around that idea:
Training uses standard tools (PyTorch) to teach a model. Veritate takes that trained model and runs it efficiently on consumer hardware. The MRI dashboard lets a human watch, in plain view, what the model is doing inside, token by token.
Three rules shape every decision:
AI today has a hardware problem. Bigger models keep pushing toward bigger GPUs, bigger data centers, and shorter useful lifespans for the machines underneath. The result is rising energy use and a steady stream of discarded equipment.
Carpathian's mission is to push back on that pattern. We research and build AI that is efficient enough to run on the hardware people already own. That has two direct effects:
Less new hardware needed. A capable CPU from the last few years is enough to run a real model. No new GPU purchase, no cloud bill, no rack of accelerators. Less e-waste. When older machines stay useful, they stay out of landfills longer. Sustainability in AI is not only about the power bill while a model runs; it is also about the equipment that gets thrown away to chase the next one.
Veritate is the working proof of that direction. It shows that careful engineering, byte-level design, and a tight performance budget can put modern AI inside reach of ordinary users, on ordinary machines, with a much smaller environmental footprint.
Veritate is the inference half of Carpathian's research effort. Together, the training side and Veritate form an end-to-end system: train a small, well-shaped model, then run it locally on hardware that already exists. The goal is simple. Make AI that is faster to run, cheaper to own, and kinder to the planet, without giving up the quality people expect.
A walkthrough of how to build semantic search with embeddings using RAG AI to index and chat with your docs, and when keyword search still wins.
Check out the Veritate Github
See the Live Stats