Choosing the right Mira model

Mira offers two models, each optimized for specific workloads. This guide helps you pick the right model.

Model comparison matrix

ModelContextMax outputSpeedCost

mira1M32KFast$

mira-thinking1M32KMedium (with thinking)$$

When to use each model

mira (Fast)

The fast model for everyday tasks. Ideal for chatbots, question answering, text summarization, translation, and code generation. The fastest and most cost-effective model.

Best for — simple Q&A, summarization, translation, short code snippets
Context — 1M tokens (~1500 pages of text)
Typical latency — < 1 second to first token

mira-thinking

A model with built-in thinking mode for complex tasks. Performs internal chain-of-thought before answering. Designed for math problems, code debugging, logical analysis, and high-accuracy tasks.

Best for — math, code debugging, logic puzzles, deep analysis
Context — 1M tokens (~1500 pages of text)
Feature — thinking mode — model reasons step-by-step before answering

Decision flowchart

Use this text-based flowchart to quickly identify the right model:

Model selection flowchart

Do you need deep reasoning / max accuracy?
   ├─ Yes → mira-thinking (includes thinking mode)
   └─ No  → mira (best cost-performance ratio)

Task-to-model matrix

Task typeRecommendationWhy

Chatbot / FAQmiraFast, cheap, accurate enough

Math problemsmira-thinkingThinking mode improves accuracy

Code reviewmira-thinkingThinking mode for deep analysis

Codebase analysismira-thinking1M context + thinking for entire projects

TranslationmiraGreat quality at low cost

Legal analysismira-thinkingLong context + high accuracy

Test generationmira-thinkingUnderstands full project context

Cost vs performance trade-offs

mira-thinking costs more due to additional thinking tokens but gives significantly more accurate results for complex tasks. For most applications, start with mira and upgrade to mira-thinking only when higher accuracy is needed.

Tip: route simple queries to mira and complex ones to mira-thinking. This reduces average cost by 40-60%.

Migrating between models

Both Mira models share the same API format. To switch between models, simply change the model parameter in your request. Prompts, tools, and system messages remain compatible.

Switching models

// Simply change the model parameter
const response = await fetch("https://api.vmira.ai/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer sk-mira-YOUR_KEY",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "mira-thinking",  // was "mira" — just change this line
    messages: [{ role: "user", content: "Analyze this document..." }],
  }),
});