Unlock exclusive introductory pricing for the newly launched Gemini 3.5 Flash.
D

DeepSeek-OCR2

Theo Yรชu cแบงu:$0.04
Ngร y phรกt hร nh:Jan 27, 2026

DeepSeek-OCR 2 is a model released by DeepSeek on January 27, 2026, using the innovative DeepEncoder V2 method, which allows AI to dynamically rearrange parts of an image based on its meaning, rather than just mechanically scanning from left to right. While maintaining high data compression efficiency, the model has achieved significant breakthroughs in multiple benchmarks and production metrics. The model can cover complex document pages with only 256 to 1120 vision tokens, achieving an overall score of 91.09% in the OmniDocBench v1.5 evaluation

Mแป›i
Sแปญ dแปฅng thฦฐฦกng mแบกi

Playground cho DeepSeek-OCR2

Khรกm phรก Playground cแปงa DeepSeek-OCR2 โ€” mรดi trฦฐแปng tฦฐฦกng tรกc ฤ‘แปƒ kiแปƒm tra mรด hรฌnh vร  chแบกy truy vแบฅn theo thแปi gian thแปฑc. Thแปญ prompts, ฤ‘iแปu chแป‰nh tham sแป‘ vร  lแบทp lแบกi ngay lแบญp tแปฉc ฤ‘แปƒ tฤƒng tแป‘c phรกt triแปƒn vร  xรกc thแปฑc cรกc trฦฐแปng hแปฃp sแปญ dแปฅng.

Technical specifications of DeepSeek-OCR-2

FieldDeepSeek-OCR-2 (published)
Release date / VersionJan 27, 2026 โ€” DeepSeek-OCR-2 (public repo / HF card).
Parameters~3 billion (3B) model (DeepSeek 3B MoE decoder + compressor).
ArchitectureVision encoder (DeepEncoder V2 / optical compression) โ†’ 3B vision-language decoder (MoE variants referenced in DeepSeek materials).
InputHigh-resolution images / scanned pages / PDFs (image formats: PNG, JPEG, multi-page PDFs via conversion pipelines).
OutputPlain text (UTF-8), structured layout metadata (bounding/flow), optional JSON K-V for downstream parsing.
Context length (effective)Uses compressed visual token sequences โ€” design goal: long, document-scale contexts (practical limits depend on compression ratio; typical pipeline yields 10ร— token reduction versus naรฏve tokenization).
Languages100+ languages / scripts (claimed multilingual coverage in product notes).

What is DeepSeek-OCR-2

DeepSeek-OCR-2 is the second major OCR/document understanding model from DeepSeek AI. Rather than treating OCR as plain character extraction, the model compresses visual document information into compact visual tokens (a process DeepSeek calls vision-text compression or its DeepEncoder family), then decodes those tokens with a 3B parameter mixture-of-experts (MoE) style VLM decoder that models text generation and layout reasoning together. The approach targets long-context documents (tables, multi-column layouts, diagrams, multilingual scripts) while reducing the sequence length and overall runtime cost compared with tokenizing every pixel/patch.

Main features of DeepSeek-OCR-2

  • Human-like reading order & layout awareness โ€” learns logical ordering of text (headingsโ†’paragraphsโ†’tables) rather than scanning fixed grids.
  • Vision-text compression โ€” compresses visual input to much shorter token sequences (10ร— typical compression target), enabling long-document contexts for the decoder.
  • Multilingual & multi-script โ€” claims support for 100+ languages and diverse scripts.
  • High throughput / self-hostable โ€” designed for on-prem inference (A100 examples), and community GGUF/local builds reported.
  • Fine-tunable โ€” repo and guides include fine-tuning instructions for domain adaptation (invoices, science papers, forms).
  • Layout + content output โ€” not just plain text: structured outputs to facilitate downstream KIE/NER and RAG pipelines.

Benchmark performance of DeepSeek-OCR-2

  • Fox benchmark / internal metric: ~97% exact-match accuracy at 10ร— compression on its Fox benchmark (the companyโ€™s benchmark focused on document fidelity under compression). This is one of the headline claims in DeepSeek marketing materials.
  • Compression trade-offs: While accuracy remains high at moderate compression (โ‰ˆ10ร—), it degrades with more aggressive compression (Tomโ€™s Hardware summarized tests showing accuracy falling to ~60% at 20ร— in some scenarios). This highlights the practical tradeoffs between throughput & fidelity.
  • Throughput: ~200k pages/day on a single NVIDIA A100 for typical workloads โ€” useful when evaluating cost/scale vs cloud OCR APIs.

Use cases & recommended deployments

  • Enterprise document ingestion & indexing: convert large corpora of annual reports, PDFs, and scanned documents into searchable text + layout metadata for RAG/LLM pipelines. (DeepSeek throughput claim is attractive for scale.)
  • Structured table extraction / financial reporting: the layout-aware encoder helps preserve table cell relationships for downstream KIE extraction and reconciliation. Validate compression level against numeric-precision needs.
  • Multilingual archive digitization: 100+ language support makes it suitable for libraries, government archives, or multinational document processing.
  • On-prem, privacy-sensitive deployments: self-hostable HF/GGUF variants enable keeping data in-house versus cloud providers.
  • Preprocessing for LLM RAG: compressing and extracting faithful text + layout for RAG ingestion where context length is a bottleneck.

How to access DeepSeek-OCR-2 via CometAPI

Step 1: Sign Up for API Key

Log in to cometapi.com. If you are not our user yet, please register first. Sign into your CometAPI console. Get the access credential API key of the interface. Click โ€œAdd Tokenโ€ at the API token in the personal center, get the token key: sk-xxxxx and submit.

cometapi-key

Step 2: Send Requests to DeepSeek-OCR-2 API

Select the โ€œdeepseek-ocr-2โ€ endpoint to send the API request and set the request body. The request method and request body are obtained from our website API doc. Our website also provides Apifox test for your convenience. Replace with your actual CometAPI key from your account. base url is Chat Completions.

Insert your question or request into the content fieldโ€”this is what the model will respond to . Process the API response to get the generated answer.

Step 3: Retrieve and Verify Results

Process the API response to get the generated answer. After processing, the API responds with the task status and output data.

Cรขu Hแปi Thฦฐแปng Gแบทp

Giรก cแบฃ cho DeepSeek-OCR2

Khรกm phรก mแปฉc giรก cแบกnh tranh cho DeepSeek-OCR2, ฤ‘ฦฐแปฃc thiแบฟt kแบฟ ฤ‘แปƒ phรน hแปฃp vแป›i nhiแปu ngรขn sรกch vร  nhu cแบงu sแปญ dแปฅng khรกc nhau. Cรกc gรณi linh hoแบกt cแปงa chรบng tรดi ฤ‘แบฃm bแบฃo bแบกn chแป‰ trแบฃ tiแปn cho nhแปฏng gรฌ bแบกn sแปญ dแปฅng, giรบp dแป… dร ng mแปŸ rแป™ng quy mรด khi yรชu cแบงu cแปงa bแบกn tฤƒng lรชn. Khรกm phรก cรกch DeepSeek-OCR2 cรณ thแปƒ nรขng cao cรกc dแปฑ รกn cแปงa bแบกn trong khi vแบซn kiแปƒm soรกt ฤ‘ฦฐแปฃc chi phรญ.

Comet Price (USD / M Tokens)Official Price (USD / M Tokens)Discount
Theo Yรชu cแบงu:$0.04
Theo Yรชu cแบงu:$0.05
-20%

Mรฃ mแบซu vร  API cho DeepSeek-OCR2

Truy cแบญp mรฃ mแบซu toร n diแป‡n vร  tร i nguyรชn API cho DeepSeek-OCR2 ฤ‘แปƒ tแป‘i ฦฐu hรณa quy trรฌnh tรญch hแปฃp cแปงa bแบกn. Tร i liแป‡u chi tiแบฟt cแปงa chรบng tรดi cung cแบฅp hฦฐแป›ng dแบซn tแปซng bฦฐแป›c, giรบp bแบกn khai thรกc toร n bแป™ tiแปm nฤƒng cแปงa DeepSeek-OCR2 trong cรกc dแปฑ รกn cแปงa mรฌnh.

#!/bin/bash

# Get your CometAPI key from https://api.cometapi.com/console/token
# Export it as: export COMETAPI_KEY="your-key-here"

curl https://api.cometapi.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -d '{
    "model": "deepseek-ocr-2",
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant."
      },
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

cURL Code Example

#!/bin/bash

# Get your CometAPI key from https://api.cometapi.com/console/token
# Export it as: export COMETAPI_KEY="your-key-here"

curl https://api.cometapi.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $COMETAPI_KEY" \
  -d '{
    "model": "deepseek-ocr-2",
    "messages": [
      {
        "role": "system",
        "content": "You are a helpful assistant."
      },
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

Python Code Example

from openai import OpenAI
import os

# Get your CometAPI key from https://api.cometapi.com/console/token, and paste it here
COMETAPI_KEY = os.environ.get("COMETAPI_KEY") or "<YOUR_COMETAPI_KEY>"
BASE_URL = "https://api.cometapi.com/v1"

client = OpenAI(base_url=BASE_URL, api_key=COMETAPI_KEY)

completion = client.chat.completions.create(
    model="deepseek-ocr-2",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"},
    ],
)

print(completion.choices[0].message.content)

JavaScript Code Example

import OpenAI from "openai";

// Get your CometAPI key from https://api.cometapi.com/console/token, and paste it here
const api_key = process.env.COMETAPI_KEY || "<YOUR_COMETAPI_KEY>";
const base_url = "https://api.cometapi.com/v1";

const openai = new OpenAI({
  apiKey: api_key,
  baseURL: base_url,
});

const completion = await openai.chat.completions.create({
  messages: [
    { role: "system", content: "You are a helpful assistant." },
    { role: "user", content: "Hello!" }
  ],
  model: "deepseek-ocr-2",
});

console.log(completion.choices[0].message.content);