Document OCR model hosted on our own GPU. Send an image plus a response_format JSON schema to extract fields. Not a conversational model — system prompts and tool calls are not supported.
DeepSeek OCR is a specialized vision model designed for high-accuracy document processing and structured data extraction. Unlike general conversational assistants, this 3.3B parameter engine focuses exclusively on interpreting images alongside provided JSON schemas to return precise field values. It delivers robust performance on both Arabic and English text, making it an ideal production-ready solution for automating invoice processing, identity verification, and form digitization within regional workflows. Note that system prompts and tool calls are unsupported, ensuring the model remains dedicated to strict extraction tasks. Developers can integrate this capability immediately via our standard API endpoint without managing complex infrastructure.
Technical efficiency is central to this model's design, operating with an 8,192 token context window at F16 quantization to balance speed and precision. Because it carries a 1x credit multiplier, it offers a cost-effective alternative to larger multimodal models for dedicated OCR tasks. Researchers will find consistent reliability in mixed-language benchmarks, particularly when extracting structured data from dense documents. Accessible from the starter tier, DeepSeek OCR allows teams to scale document intelligence workflows without prohibitive costs, ensuring seamless integration into existing pipelines while maintaining strict adherence to input schemas.
from openai import OpenAI
client = OpenAI(
base_url="https://llmapi.resayil.io/v1/",
api_key="YOUR_API_KEY"
)
response = client.chat.completions.create(
model="deepseek-ocr:latest",
messages=[
{"role": "user", "content": "Hello!"}
]
)
print(response.choices[0].message.content)const response = await fetch(
"https://llmapi.resayil.io/v1/chat/completions",
{
method: "POST",
headers: {
"Content-Type": "application/json",
"Authorization": "Bearer YOUR_API_KEY"
},
body: JSON.stringify({
model: "deepseek-ocr:latest",
messages: [
{ role: "user", content: "Hello!" }
]
})
}
);
const data = await response.json();
console.log(data.choices[0].message.content);curl https://llmapi.resayil.io/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "deepseek-ocr:latest",
"messages": [
{"role": "user", "content": "Hello!"}
]
}'