Features overview

Mira provides a comprehensive set of AI capabilities through a single API compatible with the OpenAI format. Below is an overview of all the key platform features.

Text generation

The core capability of Mira is prompt-based text generation. Mira models can write articles, code, translations, analytics, and much more. Use the /v1/chat/completions endpoint to send messages and receive responses.

Vision (image understanding)

Mira models can analyze images passed in messages. Send an image as base64 or a URL in the content array, and the model will describe, analyze, or answer questions about it. Supported formats include PNG, JPEG, WebP, and GIF.

Tool use (function calling)

Mira supports tool use — the model can request execution of functions you define. This lets you integrate the model with external APIs, databases, and services. Define a set of tools in your request, and the model will return a structured call when needed.

Extended thinking

The mira-pro and mira-max models include thinking mode with chain-of-thought reasoning to solve complex problems. The model "thinks out loud," breaking the problem into steps before giving a final answer. This significantly improves accuracy on math, logic, and programming tasks.

Streaming

Set stream: true in your request to receive the response as it is generated, token by token. This lets you display text to the user in real time, dramatically improving perceived responsiveness.

JSON mode

To get structured data, use the response_format: { type: "json_object" } parameter. The model is guaranteed to return valid JSON, simplifying parsing and integration with data processing systems.

Embeddings

Mira provides an endpoint for creating vector representations of text (embeddings). Use them for semantic search, clustering, recommendation systems, and RAG (Retrieval-Augmented Generation).

Batch processing

For processing large volumes of requests, use the batch API. Send an array of requests in a single call and receive all responses when processing is complete. This is more efficient and cost-effective for bulk operations.

Multilingual support

All Mira models are trained on multilingual data and work fluently with Russian, English, Chinese, Spanish, French, and many other languages. Models automatically detect the input language and respond in the same language.

Explore each feature

Navigate to detailed guides for each Mira capability: