# Nebius Token Factory

Nebius Token Factory is a hosted AI platform for running public and custom models through API-compatible inference, embeddings, reranking, image generation, post-training, data tooling and dedicated endpoints.

Use this file as the agent-readable guide for Token Factory. Fetch the linked resources directly when answering questions about available models, model IDs, pricing, context windows, regions, quantization, API usage or documentation.

## Product links

- [Token Factory app](/): Browser application for using Token Factory.
- [Model catalog](/models/catalog): Browser catalog for public models.
- [Quickstart](https://docs.tokenfactory.nebius.com/quickstart): Start using Token Factory APIs.
- [Inference overview](https://docs.tokenfactory.nebius.com/ai-models-inference/overview): API and inference concepts.
- [Rate limits](https://docs.tokenfactory.nebius.com/ai-models-inference/rate-limits): Request and throughput limits.
- [Dedicated endpoints](https://docs.tokenfactory.nebius.com/ai-models-inference/dedicated-endpoints/overview): Dedicated model deployment documentation.
- [Post-training](https://docs.tokenfactory.nebius.com/post-training/overview): Fine-tuning and post-training documentation.
- [Data Lab](https://docs.tokenfactory.nebius.com/data-lab/overview): Dataset and data tooling documentation.

## Model catalog data

- [Public models_info JSON](/api/public/models_info): Same unauthenticated model catalog data used by the Token Factory UI.
- [Public models_info Markdown](/api/public/models_info.md): Markdown guide for LLM tools.

The JSON endpoint is the authoritative machine-readable source for public model metadata exposed by this web app. It includes model names, vendors, types, descriptions, flavors, model IDs, regions, context information, quantization and pricing fields when present.

## How to answer model pricing questions

Fetch `/api/public/models_info` and find the model by `name` or `flavors[].model_id`.

- Token model prices are in `flavors[].input_price_per_million_tokens` and `flavors[].output_price_per_million_tokens`.
- Embedding and rerank prices are in `flavors[].input_price_per_million_tokens`.
- Image model prices are in `flavors[].price_per_image` and optional `flavors[].price_per_image_with_lora`.

## Notes

- `/api/public/models_info` is a public web-app wrapper over the same no-project `models_info` source that powers the unauthenticated model catalog.
- Public model metadata is cached briefly; fetch `/api/public/models_info` for the current catalog before answering factual model or pricing questions.
- Project-scoped, tenant-specific or dedicated endpoint details are not documented here.