Mistral AI

This page provides information about Mistral AI models available on AvalAI.

Available Models

Text and Chat Models

Mistral AI offers several powerful language models for text generation and chat applications.

mistral-large-3

The mistral-large-3 model is Mistral AI's most capable model to date, a state-of-the-art open model with a sparse mixture-of-experts architecture featuring 41B active and 675B total parameters. Released under the Apache 2.0 license.

Key Features:

• Sparse Mixture-of-Experts architecture (41B active / 675B total parameters)
• Released under Apache 2.0 license for full open-source access
• Native multimodal capabilities with image understanding
• Best-in-class multilingual performance across 40+ languages
• Achieves parity with best instruction-tuned open-weight models
• Ranks #2 in OSS non-reasoning models on LMArena (#6 overall among OSS)
• Optimized checkpoint available in NVFP4 format for efficient deployment
• Can run on single 8×A100 or 8×H100 node using vLLM

Pricing:

Input	Cached Input	Output	Per Page
$0.50/1M tokens	$0.05/1M tokens	$1.50/1M tokens	$0.001/page

Use Cases:

• Complex reasoning and analysis tasks
• Multilingual conversations and content generation
• Document understanding and processing
• Code generation and technical tasks
• Research and enterprise applications
• Agentic workflows and tool use

from openai import OpenAI

client = OpenAI(api_key="your-avalai-api-key", base_url="https://api.avalai.ir/v1")

response = client.chat.completions.create(
    model="mistral-large-3",
    messages=[
        {
            "role": "user",
            "content": "Analyze the implications of quantum computing on modern cryptography.",
        }
    ],
)

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

Responses API version This version uses `gpt-5.5` because `mistral-large-3` may not be enabled for `/v1/responses` in the current AvalAI model data.

Use this version when the selected model supports /v1/responses. messages moves to input, and the final text is read from response.output_text.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["AVALAI_API_KEY"],
    base_url="https://api.avalai.ir/v1",
)

response = client.responses.create(
    model="gpt-5.5",
    instructions="You are a helpful assistant.",
    input="Analyze the implications of quantum computing on modern cryptography.",
)

print(response.output_text)

• messages → input
• system message → instructions or a developer item
• choices[0].message.content → response.output_text
• for tools and multimodal output, inspect response.output by item type.

codestral-2501

The codestral-2501 model is a specialized 22B parameter model designed for code generation across 80+ programming languages. It sets a new standard for code generation performance and capabilities.

Key Features:

• Fluent in 80+ programming languages including Python, Java, C, C++, JavaScript, Bash, Swift, Fortran, and many more
• Completes coding functions, writes tests, and fills in partial code using a fill-in-the-middle mechanism
• 32K context window for long-range code completion and understanding
• Superior performance on benchmarks including HumanEval, MBPP, CruxEval, and RepoBench
• Excels at SQL generation and multi-language programming tasks
• Advanced fill-in-the-middle capabilities for code editing and completion

Use Cases:

• Software development and code generation
• Test creation and automation
• Code completion in IDEs and development environments
• Technical documentation generation
• Code translation between programming languages
• Debugging and code optimization

mistral-small-2503

The mistral-small-2503 model is a versatile, efficient language model designed for general text generation and understanding tasks. It offers a good balance between performance and computational efficiency.

Key Features:

• Optimized for general-purpose language tasks
• Supports a 1M token context window
• Efficient performance for routine language tasks
• Balanced capabilities across reasoning, content generation, and comprehension
• Cost-effective option for production applications

Use Cases:

• Content generation and summarization
• Question answering and information retrieval
• Text classification and analysis
• Conversational AI applications
• Document processing and understanding

OCR Models

mistral-ocr-4-0

The mistral-ocr-4-0 model is Mistral AI's latest document extraction and understanding model. It returns Markdown text together with bounding boxes, typed block classification, and inline confidence information, and supports 170 languages across 10 language groups.

mistral-ocr-latest now resolves to mistral-ocr-4-0 and uses the same pricing, so existing alias-based integrations do not need to change immediately. Use the explicit versioned ID for reproducible workflows.

Key Features:

• Returns bounding boxes for localized highlighting, citations, and redaction workflows
• Classifies typed blocks such as titles, tables, equations, and signatures
• Provides inline confidence information for verification and human review
• Preserves document structure as Markdown for RAG and indexing pipelines
• Supports 170 languages across 10 language groups
• Accepts common document formats including PDF, DOC, PPT, OpenDocument, and images
• Supports structured Document AI annotations through the same OCR endpoint
• Suitable for semantic chunking, enterprise search, form processing, invoices, and compliance workflows

Pricing:

• OCR extraction: $0.004 per page ($4 per 1,000 pages)
• Document or image annotation: $0.005 per annotated page ($5 per 1,000 pages)

Supported File Types:

• PDF documents (up to 50 MB and 1,000 pages)
• Images in PNG, JPEG, WEBP, and non-animated GIF formats

Usage

Unlike other OpenAI-compatible models, Mistral AI models use a different endpoint structure without the "/v1" path segment.

OCR Processing Examples

Using URL to Process PDF

from mistralai import Mistral

client = Mistral(server_url="https://api.avalai.ir", api_key="avalai-api-key")

document_param = {
    "type": "document_url",
    "document_url": "https://arxiv.org/pdf/1805.04770",
}

ocr_response = client.ocr.process(
    model="mistral-ocr-4-0",
    document=document_param,
    pages=list(range(0, 100)),  # Process up to 100 pages
)

print(ocr_response)

Using Base64-encoded PDF

import base64
from mistralai import Mistral

# Read and encode the PDF file
with open("document.pdf", "rb") as f:
    pdf_data = f.read()

base64_pdf = base64.b64encode(pdf_data).decode("utf-8")
document_url = f"data:application/pdf;base64,{base64_pdf}"

# Process the encoded PDF
client = Mistral(server_url="https://api.avalai.ir", api_key="avalai-api-key")

document_param = {"type": "document_url", "document_url": document_url}

ocr_response = client.ocr.process(
    model="mistral-ocr-4-0",
    document=document_param,
    pages=list(range(0, 100)),  # Process up to 100 pages
)

print(ocr_response)

Processing Specific Pages

from mistralai import Mistral

client = Mistral(server_url="https://api.avalai.ir", api_key="avalai-api-key")

document_param = {
    "type": "document_url",
    "document_url": "https://arxiv.org/pdf/1805.04770",
}

# Process only pages 0, 1, and 5
ocr_response = client.ocr.process(
    model="mistral-ocr-4-0",
    document=document_param,
    pages=[0, 1, 5],  # Only process specific pages
)

print(ocr_response)

JavaScript Example

import { Mistral } from "mistralai";

const client = new Mistral({
  apiKey: "avalai-api-key",
  baseURL: "https://api.avalai.ir",
});

const documentParam = {
  type: "document_url",
  document_url: "https://arxiv.org/pdf/1805.04770",
};

const ocrResponse = await client.ocr.process({
  model: "mistral-ocr-4-0",
  document: documentParam,
  pages: Array.from({ length: 100 }, (_, i) => i), // Process up to 100 pages
});

console.log(ocrResponse);

Document Understanding

The mistral-ocr-4-0 model can also be combined with language models to enable natural language interaction with document content. This allows you to extract information and insights from documents by asking questions in natural language.

Question Answering with Scientific Papers

from mistralai import Mistral
from mistralai.models.chat import ChatMessage

# Initialize client
client = Mistral(server_url="https://api.avalai.ir", api_key="avalai-api-key")

# Process document with OCR first
document_param = {
    "type": "document_url",
    "document_url": "https://arxiv.org/pdf/1805.04770",
}

ocr_response = client.ocr.process(
    model="mistral-ocr-4-0",
    document=document_param,
    pages=[0, 1, 2],  # Process first 3 pages
)

# Create a message with both text and document
document_content = (
    ocr_response.pages[0].text
    + "\n"
    + ocr_response.pages[1].text
    + "\n"
    + ocr_response.pages[2].text
)
message = f"""
Please analyze this scientific paper and answer questions based on its content:

{document_content}
"""

# Send the request
messages = [
    ChatMessage(role="user", content=message),
]

chat_response = client.chat(
    model="mistral-large-latest",
    messages=messages,
)

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

# Ask specific questions about the document
question = "What is the main contribution of this paper?"
messages = [
    ChatMessage(role="user", content=message),
    ChatMessage(role="assistant", content=chat_response.choices[0].message.content),
    ChatMessage(role="user", content=question),
]

chat_response = client.chat(
    model="mistral-large-latest",
    messages=messages,
)

print(f"Q: {question}\nA: {chat_response.choices[0].message.content}")

Information Extraction from Receipts

from mistralai import Mistral
from mistralai.models.chat import ChatMessage

# Initialize client
client = Mistral(server_url="https://api.avalai.ir", api_key="avalai-api-key")

# Process receipt image
document_param = {
    "type": "document_url",
    "document_url": "https://example.com/receipt.jpg",
}

ocr_response = client.ocr.process(
    model="mistral-ocr-4-0",
    document=document_param,
)

# Create a message with both text and image
message = f"""
Extract the following information from this receipt:
1. Store name
2. Date of purchase
3. Total amount
4. List of items purchased with prices

Receipt content:
{ocr_response.pages[0].text}
"""

# Send the request
messages = [
    ChatMessage(role="user", content=message),
]

chat_response = client.chat(
    model="mistral-large-latest",
    messages=messages,
)

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

Key capabilities:

• Question answering about specific document content
• Information extraction and summarization
• Document analysis and insights
• Multi-document queries and comparisons
• Context-aware responses that consider the full document
• Structured output options for downstream processing

Key Use Cases

Mistral OCR enables a wide range of document processing applications:

• Scientific Research: Convert scientific papers with complex formulas and diagrams into AI-ready formats
• Historical Preservation: Digitize historical documents and artifacts for broader accessibility
• Customer Service: Transform documentation and manuals into indexed knowledge bases
• Education: Convert lecture notes and presentations into searchable content
• Legal: Process regulatory filings and legal documents
• Engineering: Extract information from technical literature and drawings

Related Resources

• OCR 4 Announcement
• Document Processing Guide
• Processing Documents with Mistral OCR