mnc lab

mnc lab partner program · the frontier lab for world models

Build better health apps. Earn a revenue share on every integration.

Put mEVA, a self-supervised foundation world model in the H‑JEPA family from mnc lab, into the apps you build for clinics, health plans and health startups. Trained end to end on clinical data, it answers the clinical questions general models hedge on. Every integration earns you a revenue share on the API usage it brings. It runs behind an OpenAI-compatible API.

Already on the OpenAI SDK? It is a three-line change · read the docs.

the program

What partners get

A share of the revenue you bring, our newest models before anyone else, and the training and engineers to build with them.

Revenue share

Earn a share of the API revenue from every app you build mEVA into, and every client you resell it to.

Closed-beta access to new models

Use our new models in closed beta, before release, so your clients get them first.

Training and forward deployed engineers

We train your team, and our engineers build client projects with you, from scoping to go-live.

built for
App developers and studiosSystems integratorsConsultancies
benchmark

Better clinical accuracy than GPT-6 Astra, for 5× less

Benchmarks test exam questions; real-world care, the 360° patient view below, is where mEVA excels. On closed-form accuracy across the six closed-answer formats of each ReMedQA question, in our own evaluation, mEVA scores 89.56% against GPT-6 Astra’s 87.39%.

mEVA
89.56%
Claude Opus 5.5
89.38%
GPT-6 Astra
87.39%
Best open-weights
85.58%
GPT-5.6 Sol
73.81%
usd · per 1M tokens
input
output
mEVA
$2
$10
GPT-6 Astra
$10
$50
Claude Opus 5.5
$4
$20
GPT-5.6 Sol
$4
$20

One question, seven versions

ReMedQA rewrites medical exam questions, including the USMLE (the US physicians’ bar exam), seven ways: answers relabelled or moved, and a “none of these” option added. The score averages the six closed-answer formats.

Shortcuts stop working

On a plain multiple-choice test a model can score by spotting how answers are written. Here those patterns no longer help.

Run it yourself

Run the benchmark on mEVA yourself at platform.moonco.one. The harness is also public under MIT at github.com/moonco-one/remedqa-benchmark, to run beside any model you already use.

ReMedQA paper · benchmark harness · prices are public list prices per 1M tokens.

where mEVA excels

Real-world care needs a 360° patient view

mEVA analyzes images, video, documents and sensor data in one model, and returns output a care team can act on.

Wounds & photos

  • Tissue classification
  • Wound area measurement
  • Progression analysis over time

Motion & video

  • Range of motion
  • Stroke facial asymmetry
  • Mobility assessment

Medical documents

  • OCR of paper charts
  • Extraction from lab photos
  • Structuring clinical notes

Sensor data

  • Continuous vital signs
  • Smart wearable telemetry
  • Biometric data streams
clinical areas
Wound careOrthopedicsStroke & neuroDocs & chartsRemote monitoring
use cases

What you can build for clients

The projects healthcare clients already ask for, on one model and one API.

B2C health apps

Symptom checks, photo analysis and guided recovery inside your consumer app.

  • Symptom checks from photos and short questions
  • Wound and skin photo analysis
  • Guided recovery plans between visits

Patient engagement

Clinical answers and escalation inside the chat and voice agents patients already use.

  • Clinical answers between visits
  • Escalation to a clinician, with the reason attached
  • Follow-up outreach after discharge

Telehealth

Pre-visit intake, plus photo assessment between virtual visits.

  • Pre-visit intake
  • Photo assessment between visits
  • Follow-up triage

Remote monitoring

Device and wearable readings turned into flags a care team can act on between visits.

  • Vital-sign and wearable trend analysis
  • Deterioration flags for the care team
  • Monthly monitoring summaries

EHRs

Chart extraction, patient summaries and risk flags inside the clinician’s workflow.

  • Chart extraction from scanned documents
  • Patient summaries before each visit
  • Risk flags inside the clinician’s workflow

Records digitization

Archived paper and scanned records turned into structured, searchable patient histories.

  • Extraction from legacy and scanned records
  • Patient-history summaries across thousands of pages
  • Structured output ready for EHR import

RCM

Structured data from charts and notes to support documentation and billing.

  • Structured data from charts and notes
  • Documentation support
  • Billing-ready outputs

Insurance

Faster record review and risk stratification across member populations.

  • Faster record review
  • Risk stratification across members
  • Evidence attached to every flag
built to move
Fewer 30/60/90-day readmissionsFewer avoidable ED visitsFewer unplanned acute escalationsHigher follow-up completionFaster response to deteriorationLower cost per episode
the switch

Three changes. The rest of your code stays.

If your app calls Chat Completions through the OpenAI SDK, change the key, the base URL and the model name.

python

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["OPENAI_API_KEY"],
    api_key=os.environ["MOONCO_API_KEY"],
    base_url="https://api.moonco.one/v1",
)

reply = client.chat.completions.create(
    model="<your OpenAI model>",
    model="moon-1",  # mEVA
    messages=messages,
)

typescript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  apiKey: process.env.MOONCO_API_KEY,
  baseURL: "https://api.moonco.one/v1",
});

const reply = await client.chat.completions.create({
  model: "<your OpenAI model>",
  model: "moon-1", // mEVA
  messages,
});

Move the clinical calls, keep the rest. Start with one code path, such as intake, triage or a chart summary, and send it to mEVA. Embeddings, speech and image generation stay where they are.

the cost

What your clients would pay on mEVA

on GPT-6 Astra
$2,000 a month
on mEVA
$400 a month
you keep
$1,600 a month · $19,200 a year

List prices per 1M tokens: mEVA $2 / $10, GPT-6 Astra $10 / $50, GPT-5.6 Sol $4 / $20. OpenAI’s cached-input and batch discounts are not included.

build or resell

Build on the API, resell it, or both

Most partners start with one and add the other.

Build on the API

Put mEVA inside the solutions you deliver: patient apps, clinical workflows, EHR integrations.

  • Call mEVA from client products through one API
  • Bill your own discovery, integration and support work
  • You own the code and the product you build

Resell API access

Offer mEVA to your clients as part of your portfolio and earn on the usage you bring.

  • Register deals so the account stays yours
  • Sales materials and joint calls with our team
  • We run the model, hosting and updates
getting started

How to get started

  1. Get a key

    Write to business@moonco.io for an invitation to platform.moonco.one, then top up from $10.

  2. Switch one call

    Point one clinical code path at mEVA and run your existing test prompts against both models.

  3. Ship a first client project

    Put it in front of a client, then apply to the partner program.

  4. Grow together

    Take mEVA to more clients and co-sell on new accounts.

faq

Questions from partners

Can we send patient data?

Yes. mEVA runs in a HIPAA-compliant cloud, and we sign a Business Associate Agreement with you before protected health information reaches the API. Write to business@moonco.io.

Do you train on our clients’ data?

No, not by default. Request content is not archived by default, and client data is used for training only with your prior, explicit consent.

Can it run on our client’s infrastructure?

Yes. mEVA deploys to any major cloud, a private cloud or on-premises, so data is stored and processed in the country where local law requires. We agree the setup for each project.

Who owns the client relationship?

You do. You lead the engagement and the contract; we support you in the background and join calls when you ask us to.

Is mEVA cleared as a medical device?

mEVA is built to support clinicians, not to replace their judgement. Whether a product you build needs regulatory clearance depends on its intended use and the country it runs in. We share documentation and evidence to support your assessment.

apply

Join the mnc lab partner program

Tell us about your company, your healthcare clients and the app you would move first. We will reply with a key, program terms and next steps.

  • A revenue share on every integration
  • Closed-beta access to new models
  • Training and forward deployed engineers

or write to business@moonco.io

Even the best healthcare AI, ours included, is not yet capable of replacing clinical judgment. The model is built to be checked by a clinician, not to stand in for one.