Modelspage

DeepSeek OCR 2

High-accuracy OCR model for extracting text from documents, screenshots, receipts, and natural scenes

deepseek/deepseek-ocr-2
Lab
DeepSeek
Family
-
Providers
3
Context
8,192
Output limit
8,192
Knowledge
-
Release
2026-01-27
Updated
2026-01-27
Weights
Open
Input
Output types
Capabilities
temperature

Providers

3
Providers
Novita AIDeepSeekdeepseek/deepseek-ocr-28,1928,192$0.03 / $0.03NoNo-Yes
Privatemode AIDeepSeekdeepseek-ocr-28,1928,192$0.8897 / $1.4675NoNo-Yes
Regolo AIDeepSeekdeepseek-ocr-24,0004,000$0.00 / $0.00NoNo-Yes
No rows match the current search.

No rows match the current search.

Menu

Esc

Search

No matching results.

How to use

Modelspage lets you browse models, compare providers, and explore rankings. Open a model for its specifications and provider prices.

API

You can access provider data, provider-agnostic model metadata, or the combined catalog through JSON endpoints. Specialized model types are omitted by default; the site includes all model types.

Logos

Provider logos are available at /logos/{provider}.svg where {provider} is the provider ID. Lab logos are available at /logos/labs/{lab}.svg.

Upstream SDK

The models.dev SDK provides typed catalog queries and upstream snapshots for offline use.

Contribute

The model catalog is stored in GitHub as TOML files organized by provider and canonical model.