Straight answers.

Eigna builds a company its own AI. We set it up where the company decides, keep teaching it the company’s work, and release a new version only when it is better. The company keeps the model.

Companies first: enterprises, small and mid-size companies, AI-first startups with a heavy token bill, and the public sector and critical infrastructure. Then people and homes, with Eigna Box Personal and Home, which are planned.

The first expert, the scoring harness and sealed tests, training on local machines and H200 GPUs, and benchmarks.eigna.ai. The Console and the Model Deed are in development. Learning in production, the run modes as a product, Eigna Companion, the first Eigna Box and connectors are planned.

Keep them for email and office work. For the core work, on data that cannot leave, a rented assistant is generic, priced per seat and changed on someone else’s schedule. Eigna takes that core work, and you own the result.

Eigna Labs, Inc., formed in Delaware in October 2026 by Omar Garcia, Andrés Soltermann, James Sherborne and Roman Fritschi. Engineers first, evidence by habit.

An open base model plus the work your people approve makes a model that belongs to you. Every model comes with a Model Deed: the owner, the version, the date and the fingerprint of the weights, signed.

Your people approve the work. A new version is trained on it, tested on a sealed set it has never seen, and released only when it beats the one you use today. A correction is not training data until someone approves it. Learning in production is planned.

It does not go live. A tie keeps the current version, and you can roll back to the previous release in 60 seconds, at any time.

Yes. The weights, the test set and the release record are yours, with a Model Deed. You stay for the next release, not because you are locked in.

We hope so. Better open models lower our cost, and we re-base on the best one, through the gate. What does not become a commodity is your data, your sealed test set, your release record and our Formula for your field.

With Eigna Local, your data stays on your machines: data never leaves. With Eigna Hybrid and Eigna Cloud, data stays in the region you choose.

It learns only from work you approve. With Local, an encrypted training snapshot goes to a GPU we rent in a region you approve, and is deleted after the run. Every new version must beat the current one on a sealed test set.

No. Client data never moves between clients. Only examples you approve train your model.

You do. In Eigna Cloud, your data sits on your own volumes, with keys you hold.

Every release, its test result and who approved it. A full audit log of every request is planned.

None yet, and we will not suggest otherwise. Ask for our security questionnaire and we answer it in writing, per run mode, with what is built and what is planned.

Today, Eigna Local runs on a closed network. An air-gapped profile is planned.

Eigna Local runs on your servers or an Eigna Box on your site. Eigna Hybrid runs small experts on your machine and the main expert on a GPU in your region. Eigna Cloud runs on a dedicated GPU in the region you choose, run by us.

Where the data must stay decides it. If it must never leave your site, Local. If it must stay in your country or region, Hybrid. If your cloud is fine and your use is heavy and steady, Cloud. If you are not sure, we decide it together on the call.

A Mac, an NVIDIA workstation or a GB10 box with about 32 GB of memory or more, or an Eigna Box. The first Eigna Box is planned, and we build it ourselves.

Moving between modes is planned.

The three run modes as a product are planned. Your first model is set up with you during the Test Lab.

Per site or per GPU, never per seat. A setup fee once, then a fixed monthly fee to run the model and keep it learning. No token bill. All prices are proposed until they are in your offer.

Setup $3,250 a site for Local and Hybrid, $5,500 a dedicated GPU for Cloud. Running $1,100 a site a month for Local, $1,900 for Hybrid, and $7,700 a dedicated GPU a month for Cloud on a 12-month term. Proposed, not yet tested with buyers.

For heavy, steady use of mid and top tier models. One Local site at $1,100 a month costs about the same as 275 million Claude Sonnet 5.5 tokens a month at list prices. Against light use of the cheapest models, renting stays cheaper, and we make no cost claim there.

We make no cost claim against light use. The flat fee wins for heavy, steady use of the mid and top tiers; for the rest, the case is quality on the task and control of every version.

$4,900 for four weeks on one task, proposed, credited against setup when you continue.

One task, your own data, a sealed setup and your baseline, in about four weeks. We show whether an Eigna expert beats your baseline.

You keep the report: what we tried, what it scored and why it fell short.

The test set becomes your first release gate, and the fee is credited against setup.

One task, its data, today’s baseline if you have one, one person to score the work, and a start date.

The Eigna team running your lab, inside the sealed setup agreed with you. Partners never see client data unless you ask them to, and your data never trains anyone else’s model.

One team, one task and one Test Lab first. Designed for a signed installer, keys held by you, a Model Deed and rollback in 60 seconds. Cleared partners where the work requires them.

Technology partners (model makers, GPU and cloud platforms, software vendors, data centres), distribution partners (consultancies, integrators, resellers, telecoms) and hardware partners (device and server makers, contract manufacturers, retail).

Six roles, from light to deep: refer, resell, deliver, host, embed and co-build. Whoever sells, Eigna runs the models, the loop and the proof.

The partner keeps the client and the relationship.

Never, unless the client asks them to.

Exclusivity is earned with results, never given up front.

Not yet. Everything we describe is open to the first partners, and they shape the programme with us.

Your AI infrastructure and tech partner, and an early investor. We run your own model and its infrastructure, keep it learning on your traffic, and take equity early when the venture fits.

Companies whose model is their business: the model is the product, token use is heavy and growing, or a wrong or leaked answer is costly. If two are true, we should talk.

Only a test on your task can say. The Test Lab measures it against your current model first. If it does not win, you keep the report.

Only when it fits both sides. Compute and engineering can be paid in cash or against equity. Every term is set with counsel. No promise of returns.

No. Your work trains only your model, and you keep the weights.

Send one task and the last three months of your token bill. We answer within a week with a fixed-cost plan and a Test Lab proposal.