MatData

Products

Models & Datasets

Checkout is open. A purchase creates a portal immediately. Weight dumps and large datasets download after engineering stages them in R2.

AI Models

Atlas Full Weights

Flagship language model. Own the weights.

Flagship

The complete Atlas model weight package. Includes all checkpoint files, tokenizer configs, and a starter inference script. Released under a commercial use license. Ready for fine tuning, deployment, or research.

Architecture

Transformer (decoder only)

License

Commercial

Format

safetensors / GGUF / ONNX

Size

~18 GB

  • Commercial use license included
  • All checkpoint files
  • Tokenizer + config files
  • Starter inference script
  • 30 days of email support

$499

paid once

Atlas Inference License

Deploy Atlas in your product under a one year license.

Popular

A 12 month inference license for Atlas. Receive the weights and a legal right to serve Atlas in commercial products. Includes quarterly weight updates.

License Duration

12 months

Updates

Quarterly

Support

Priority (email + chat)

Deployment

Your hardware or the cloud

  • 12 month commercial inference license
  • Quarterly weight updates included
  • Priority email & chat support
  • Usage in SaaS/internal products allowed
  • Unlimited inference calls

$999

paid once

Chronos

In development

Our second model is being trained with a different objective. It is not for sale. Specs will land here when we publish them.

Synthetic Datasets

Synthetic Dataset Core

500 GB of high fidelity synthetic training data.

500 GB· JSONL / Parquet / CSV

Our Core synthetic dataset bundle: 500 GB across 20+ categories including instruction following pairs, code generation, reasoning chains, and conversational dialogue. Fully formatted for direct fine tuning.

Size

~500 GB

Format

JSONL / Parquet / CSV

Categories

20+

License

Commercial training use

  • 500 GB of synthetic training data
  • 20+ data categories
  • No PII, ready for audit
  • Train/val splits included
  • Data cards & documentation

$199

paid once

Synthetic Dataset Pro

3 TB of curated synthetic data across 50+ categories.

Best Value
3 TB· JSONL / Parquet / CSV

The Pro tier gives you everything in Core plus 2.5 additional terabytes: multi turn dialogue, domain specific technical corpora (legal, medical, finance), and multilingual instruction data across 12 languages.

Size

~3 TB

Format

JSONL / Parquet / CSV

Categories

50+

Languages

12

  • Everything in Core
  • 3 TB total, 50+ categories
  • Domain specific corpora (legal, medical, finance)
  • Multilingual data (12 languages)
  • Versioned access, with updates for 6 months

$699

paid once

Bundles

Save $590

Atlas + Pro Data Bundle

The complete package: model weights plus 3 TB of data.

Get Atlas Full Weights and the Pro Synthetic Dataset together at a significant discount. Everything you need to fine tune, adapt, and deploy your own production grade language model.

  • Atlas Full Weights (commercial license)
  • Pro Synthetic Dataset (3 TB)
  • Dedicated onboarding session (1 hr)
  • 6 months of data updates
  • Priority support

$1,099

paid once

Need a Custom Deal?

Volume discounts, custom licensing, white label agreements, or enterprise work. Reach out and we will work something out.

Contact Sales