> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/cheahjs/free-llm-api-resources/llms.txt
> Use this file to discover all available pages before exploring further.

# Choosing a Provider

> Compare LLM API providers and find the best fit for your use case

## Overview

With 26+ free and trial-credit LLM providers available, choosing the right one depends on your specific needs. This guide helps you compare providers based on key factors like rate limits, model selection, and use cases.

## Quick Comparison

### By Rate Limits

<AccordionGroup>
  <Accordion title="High Request Volume (10,000+ requests/day)">
    **Best Options:**

    * **Google AI Studio** - Up to 14,400 requests/day for Gemma models
    * **Cerebras** - 14,400 requests/day with 1M tokens/day
    * **Groq** - Up to 14,400 requests/day for Llama models
    * **GitHub Models** - Varies by Copilot tier

    <Tip>These providers are ideal for production applications with consistent traffic.</Tip>
  </Accordion>

  <Accordion title="High Token Volume (500K+ tokens/day)">
    **Best Options:**

    * **Mistral La Plateforme** - 500K tokens/minute, 1B tokens/month per model
    * **Cerebras** - 1M tokens/day across all models
    * **Google AI Studio** - 250K tokens/minute for Gemini models
    * **Groq** - Up to 70K tokens/minute for Compound models

    <Info>Perfect for applications that need to process large documents or generate long-form content.</Info>
  </Accordion>

  <Accordion title="Moderate Usage (100-1000 requests/day)">
    **Best Options:**

    * **OpenRouter** - 50 requests/day (1000 with \$10 lifetime topup)
    * **Groq** - 250-1000 requests/day depending on model
    * **Cohere** - 1000 requests/month shared across models
    * **Mistral Codestral** - 2000 requests/day

    <Note>Suitable for development, prototyping, and small-scale applications.</Note>
  </Accordion>
</AccordionGroup>

### By Model Selection

<Tabs>
  <Tab title="Latest Frontier Models">
    **GitHub Models** offers the most cutting-edge models:

    * GPT-5, GPT-5-mini, GPT-5-nano
    * OpenAI o3, o3-mini, o4-mini
    * Grok 3, Grok 3 Mini
    * DeepSeek-R1, DeepSeek-V3
    * Llama 4 Maverick and Scout

    <Warning>GitHub Models has extremely restrictive input/output token limits. Not suitable for production.</Warning>
  </Tab>

  <Tab title="Large Open Models (70B+)">
    **Best for reasoning and complex tasks:**

    * **OpenRouter** - Llama 3.3 70B, Hermes 3 405B, Qwen 3 Next 80B
    * **Cerebras** - Qwen 3 235B, Llama 3.3 70B
    * **Groq** - Llama 3.3 70B (ultra-fast inference)
    * **Hyperbolic** - DeepSeek V3, Llama 3.1 405B (trial credit)
  </Tab>

  <Tab title="Code Generation">
    **Specialized for coding tasks:**

    * **Mistral Codestral** - Dedicated code model with 2000 req/day
    * **OpenRouter** - Qwen 3 Coder models
    * **Cloudflare Workers AI** - Qwen 2.5 Coder, DeepSeek Coder
    * **GitHub Models** - Codestral, various code-focused models
  </Tab>

  <Tab title="Multimodal (Vision)">
    **Image understanding capabilities:**

    * **Google AI Studio** - Gemini 2.5 Flash, Gemini 3 Flash
    * **GitHub Models** - Llama 3.2 90B Vision, GPT-4o
    * **Groq** - Llama Guard 4 (visual content moderation)
    * **Cloudflare** - Llama 3.2 11B Vision
  </Tab>

  <Tab title="Multilingual">
    **Strong non-English support:**

    * **Cohere** - Aya family (multilingual specialist)
    * **Alibaba Cloud** - Qwen models (Chinese focus)
    * **OpenRouter** - Qwen 3 models
    * **Groq** - Kimi K2 models
  </Tab>
</Tabs>

## Use Case Recommendations

<CardGroup cols={2}>
  <Card title="Chatbots & Assistants" icon="comments">
    **Recommended Providers:**

    * **Groq** - Ultra-fast inference for real-time chat
    * **Google AI Studio** - Generous Gemini quotas
    * **OpenRouter** - Variety of personality-tuned models

    **Key Factors:** Low latency, high requests/day, conversational models
  </Card>

  <Card title="Content Generation" icon="pen-to-square">
    **Recommended Providers:**

    * **Mistral La Plateforme** - 1B tokens/month
    * **Cerebras** - High token limits
    * **Google AI Studio** - 250K tokens/minute

    **Key Factors:** High token throughput, long context windows
  </Card>

  <Card title="Code Development" icon="code">
    **Recommended Providers:**

    * **Mistral Codestral** - Specialized code model
    * **OpenRouter** - Qwen 3 Coder
    * **GitHub Models** - Latest code models

    **Key Factors:** Code-specific training, high accuracy
  </Card>

  <Card title="Research & Analysis" icon="flask">
    **Recommended Providers:**

    * **OpenRouter** - Access to Hermes 3 405B
    * **Cerebras** - Qwen 3 235B
    * **GitHub Models** - o3, DeepSeek-R1

    **Key Factors:** Large parameter counts, reasoning capabilities
  </Card>

  <Card title="Prototyping" icon="wrench">
    **Recommended Providers:**

    * **OpenRouter** - Easy start, multiple models
    * **Groq** - Fast iteration cycles
    * **Vercel AI Gateway** - Multi-provider routing

    **Key Factors:** Easy setup, flexibility, good documentation
  </Card>

  <Card title="Production Apps" icon="server">
    **Recommended Providers:**

    * **Google AI Studio** - Reliable, high quotas
    * **Cerebras** - Consistent performance
    * **Mistral La Plateforme** - High monthly limits

    **Key Factors:** Reliability, high daily limits, SLA
  </Card>
</CardGroup>

## Decision Matrix

| Priority           | Best Providers                          | Notes                                  |
| ------------------ | --------------------------------------- | -------------------------------------- |
| **Speed**          | Groq, Cerebras                          | Both offer ultra-fast inference        |
| **Variety**        | OpenRouter, GitHub Models               | Access to dozens of models             |
| **Reliability**    | Google AI Studio, Mistral               | Established platforms with SLAs        |
| **Privacy**        | Google AI Studio (EU), HuggingFace      | Data not used for training             |
| **Free Tier Size** | Mistral La Plateforme, Cerebras         | Highest token quotas                   |
| **Trial Credits**  | Baseten ($30), Modal ($30), AI21 (\$10) | One-time credits for testing           |
| **No Signup**      | None                                    | All providers require account creation |
| **No Phone**       | OpenRouter, Cerebras, Cohere            | No phone verification                  |

## Special Considerations

<AccordionGroup>
  <Accordion title="Data Privacy">
    **Providers that do NOT use your data for training:**

    * Google AI Studio (UK/CH/EEA/EU only)
    * HuggingFace (depends on model provider)

    **Providers requiring data training opt-in:**

    * Mistral La Plateforme (free tier only)
    * Google AI Studio (outside EU/UK/EEA/CH)

    <Warning>Always review the privacy policy and terms of service for production deployments.</Warning>
  </Accordion>

  <Accordion title="Phone Verification">
    **Requires phone number:**

    * NVIDIA NIM
    * Mistral La Plateforme
    * Mistral Codestral
    * NLP Cloud

    **No phone required:**

    * OpenRouter
    * Groq
    * Cohere
    * Cerebras
    * GitHub Models (requires GitHub account)
  </Accordion>

  <Accordion title="Geographic Restrictions">
    **Google Cloud Vertex AI:**

    * Very stringent payment verification
    * May not be available in all regions

    **Alibaba Cloud:**

    * International version for non-China users

    **Scaleway:**

    * European infrastructure (France)

    <Info>Most providers are globally accessible, but check regional availability for compliance.</Info>
  </Accordion>

  <Accordion title="Context Window Limits">
    **NVIDIA NIM:**

    * Models tend to be context window limited

    **GitHub Models:**

    * Extremely restrictive input/output token limits

    **Best for long context:**

    * Mistral La Plateforme (depends on model)
    * Google AI Studio (Gemini models)
    * OpenRouter (varies by model)
  </Accordion>
</AccordionGroup>

## Multi-Provider Strategy

For production applications, consider using multiple providers:

<Steps>
  <Step title="Primary Provider">
    Choose based on your main use case (e.g., Cerebras for high volume, Groq for speed)
  </Step>

  <Step title="Backup Provider">
    Select a second provider with similar models for failover (e.g., OpenRouter or Google AI Studio)
  </Step>

  <Step title="Router Implementation">
    Use Vercel AI Gateway or implement your own routing logic to distribute requests
  </Step>

  <Step title="Monitor Usage">
    Track rate limits across providers and switch when approaching limits
  </Step>
</Steps>

<Tip>
  **Pro Tip:** Start with OpenRouter for prototyping (easy setup, multiple models), then optimize for specific providers in production.
</Tip>

## Next Steps

<CardGroup cols={2}>
  <Card title="Free Providers" icon="infinity" href="/providers/free/openrouter">
    Explore detailed documentation for all 13 always-free providers
  </Card>

  <Card title="Rate Limits Guide" icon="gauge" href="/guides/rate-limits">
    Learn how to optimize and track your rate limit usage
  </Card>
</CardGroup>