6 AI Models, 6 Use Cases: A Developer's Guide to Choosing the Right One

When you're building with LLMs, there's no single "best" model. The best model depends entirely on what you're building.

Here's a practical guide to which model fits which job — with real code examples and cost breakdowns.

The 6 Model Types

⚡ Fast — General Chat & Text

Best for: Customer support chatbots, content generation, summarization

This is your workhorse. Fast responses, lowest cost. Use it for the 80% of requests that don't need deep reasoning.

const completion = await client.chat.completions.create({
  model: 'dubhe-fast', // $0.30/M input
  messages: [{ role: 'user', content: 'Write a product description' }]
});

Cost per request: ~$0.00022
When to use: Default for all simple queries

💻 Code — Programming Tasks

Best for: Code generation, debugging, code review

Specialized for programming. Handles complex code patterns and multiple languages.

const completion = await client.chat.completions.create({
  model: 'dubhe-code', // $0.80/M input
  messages: [{ role: 'user', content: 'Write a React hook that debounces API calls' }]
});

🤖 Agent — Automation

Best for: Function calling, multi-step workflows, tool use

Optimized for agentic patterns — structured outputs, tool calling, and complex instructions.

⊕ Plus — Long Context

Best for: Document analysis, full-codebase review, long conversations

Handles up to 1M tokens of context.

👁 Vision — Image Understanding

Best for: Image analysis, OCR, screenshot understanding

const completion = await client.chat.completions.create({
  model: 'dubhe-vision',
  messages: [
    { role: 'user', content: [
      { type: 'text', text: 'What error is shown?' },
      { type: 'image_url', image_url: { url: 'https://example.com/error.png' } }
    ]}
  ]
});

🧠 Reasoner — Deep Reasoning

Best for: Complex problem-solving, math, logic, analysis

Spends extra compute tokens on "thinking" before answering.

Decision Tree

Is the task code-related?
  → Yes → Use Code model
  → No → Does it need images?
    → Yes → Use Vision model
    → No → Complex reasoning?
      → Yes → Use Reasoner model
      → No → Long context needed?
        → Yes → Use Plus model
        → No → Agentic (tool use)?
          → Yes → Use Agent model
          → No → Use Fast model

Cost Strategy

1. Route simple queries to Fast model (saves 10-20x)
2. Route coding questions to Code model
3. Fall back to Reasoner only when uncertain
4. Use Vision only when images are submitted

This approach can cut API costs by 60-80%.

Try It Yourself

All these models are available through a single API endpoint at Dubhe Hub — just change the model name in your code.

Pricing starts at $0.30/M input tokens. Free tier: 100K tokens to test before committing.