Small AI.
Serious research.

We build and study efficient language models designed to push the limits of what can be done with limited compute.

Explore our models Read our research

Horizon / Research artifactRecord 01

Finding

Research output is easier to verify when the conclusion, supporting material, and uncertainty remain distinct.

Evidence

Keep findings separate from supporting evidence.
State uncertainty instead of smoothing it away.
List only sources actually available to the model.

Uncertainty

Quality still depends on the selected model, supplied context, retrieval coverage, and the specificity of the question.

Sources

No external sources were supplied in this example. Verify independently.

See the method

Trusted by employees of

Affiliations indicate individual users, not formal company endorsements.

North ML / research practice

Built without hyperscale compute.

Efficient systems

North ML experiments with model architecture, training, distillation, fine-tuning, and inference — with an emphasis on models ordinary hardware can actually run.

Research, shared carefully

Models. Datasets. Experiments. Failures. We share work openly when it makes sense, while keeping some systems and research closed until North ML is ready to release them.

Our goal

Make capable AI smaller, faster, and more accessible.

Working record

North shows the constraints with the product.

Systems noteSmall Models, Serious Research Economics MethodEvaluation Before Eloquence Model recordNorth ML on Hugging Face

Horizon

Bring the hard question.

Get the finding, the uncertainty, and the next thing worth verifying.

Request access

New profiles join the waitlist after email verification.

Small AI.Serious research.