Frequently asked questions

Local AI means running AI models directly on your own computer or servers instead of calling a cloud API. Your prompts, files, and the model itself stay on hardware you control. Icosa Computing builds local AI products, including Zeno for macOS and LM Shop for training custom models, so teams get frontier-level AI without sending data to a third party.

The inference is. Cloud AI charges per token, so heavy use means growing bills. With local AI, your computer does the work, so usage costs nothing beyond electricity. An agent can run thousands of steps overnight for pennies. That is as close to free AI as it gets: no subscriptions scaling with usage, no token meters, no rate limits.

Private AI keeps everything on machines you control: prompts, documents, and model weights. A cloud chatbot sends your data to someone else's servers, where it can be logged, indexed, or used for training. With Icosa's private AI, your files never leave your computer, which is why firms in finance, law, and healthcare choose it.

Secure AI removes the attack surface that comes with sending data out. Because Icosa's models run on-device or on-premises, there is no third-party data processor, no cloud breach exposure, and no vendor that can change terms or shut down access. Compliance teams can audit exactly where data lives: on your own hardware.

On-prem (on-premises) AI is AI deployed inside your own infrastructure, such as a workstation, an office server, or your own data center, rather than rented from a cloud provider. Icosa specializes in on-prem AI for regulated firms: small, domain-expert models that run on a few GPUs or even a MacBook, instead of frontier-scale models that need a 1.5 TB cluster.

An SLM is a language model compact enough to run on local hardware, typically billions of parameters instead of trillions. A general frontier model spends most of its capacity on knowledge your team never uses. An SLM fine-tuned on your domain matches frontier output on the work that matters, while being faster, cheaper, and fully private.

Combinatorial Reasoning is Icosa's physics-based optimization method, developed with NASA and USRA and funded by $1.4M from the U.S. National Science Foundation. Instead of accepting a model's first chain of thought, it samples many candidate reasoning paths, maps selection to an optimization problem, and picks the strongest subset before answering. This is how small local models reach reasoning quality associated with much larger hosted systems.

On its trained domain, yes. Open-weight models now catch frontier releases within weeks, and Icosa narrows what remains through fine-tuning on your data, distillation, and Combinatorial Reasoning. A model trained on your contracts catches the same clause-level risk a frontier model would, because it was built on your work.

Icosa Computing is a New York-based deep-tech company building local, private AI. Backed by NSF research funding and collaborations with NASA, Fujitsu, NEC, and Toshiba, Icosa turns open models into domain experts that run entirely on hardware you own, with zero token costs.

Zeno, Icosa's local AI workspace for macOS launching August 19, 2026, runs entirely offline. It learns your documents, code, and terminology on your machine, with no cloud dependency, no meter, and no data leaving your Mac. There is nothing to configure and no model selection to manage.

Upload your documents to LM Shop, Icosa's no-code training platform, and it produces a custom small language model in minutes. Download the model and run it locally with Zeno, llama.cpp, or any compatible app. You own the resulting model outright. It is yours to keep and deploy.

For individuals, an Apple Silicon Mac. A standard MacBook is enough, because Icosa's models are optimized for local inference. For teams deploying on-premises at larger scale, a few GPUs replace what would otherwise require a frontier-scale cluster and a dedicated infrastructure team.

Apply for early access to Zeno ahead of the August 19 launch, or try LM Shop today to train and download your first custom model. Teams that want a walkthrough can request a demo with the Icosa team.