Skip to content
Rifx.Online Docs Rifx.Online Docs Rifx.Online Docs

Models

Rifx.online offers a diverse model library covering various artificial intelligence applications. We integrate models from top AI providers such as OpenAI, Anthropic, Meta, Google, and others, providing access through a unified API interface.

Main Model Categories

1. Text Processing Models (Text-to-Text)

Currently available main models include:

  • Ministral 8B

    • Features: Unique interlaced sliding window attention pattern
    • Context: 125K
    • Price: $0.1/M input tokens, $0.1/M output tokens
    • Suitable for: Edge computing scenarios

  • Qwen2.5 Series

    • 7B version: Suitable for general application scenarios
    • 72B version: Provides stronger coding and reasoning capabilities
    • Context: 128K
    • Features: Richer knowledge base compared to Qwen2

  • Claude 3.5 Sonnet

    • Provider: Anthropic
    • Context: 195.31K
    • Price: $3/M input tokens, $15/M output tokens
    • Specialties: Code writing, editing, and execution

2. Multimodal Models (Text-Image-to-Text)

Models supporting both text and image inputs:

  • Meta Llama 3.2 Vision Series

    • 11B version: Suitable for general vision tasks
    • 90B version: Suitable for complex visual reasoning
    • Context: 128K
    • Image processing fee: $0.079-0.506/K image tokens

  • Qwen2-VL 72B

    • Features: Supports image understanding at various resolutions and ratios
    • Context: 32K
    • Price: $0.4/M tokens (input/output), $0.578/K image tokens

3. Professional Models

  • OpenAI o1 Series
    • mini version: $3/M input, $12/M output
    • preview version: $15/M input, $60/M output
    • Features: Focused on STEM fields including mathematics, science, and programming

Model Selection Guide

Consider the following factors when choosing models:

1. Performance Requirements

  • Response Speed: For quick responses, choose models with smaller parameters like Llama 3.2 1B
  • Accuracy: For high accuracy requirements, consider large parameter models like Qwen2.5 72B

2. Cost Considerations

  • Low-cost Solutions:
    • Llama 3.2 1B: $0.01/M input, $0.02/M output
    • Qwen2 7B: $0.054/M tokens
  • High-performance Solutions:
    • Claude 3.5 Sonnet
    • OpenAI o1 series

3. Special Function Requirements

  • Long Text Processing: Choose models with larger context windows
  • Image Processing: Use multimodal models like Llama Vision or Qwen2-VL
  • Professional Fields: Select specially optimized models based on specific scenarios

For more detailed information, visit Rifx.online Models.