Skip to main content

​
Overview

RAG Chat supports multiple OpenAI GPT models, each with different capabilities, speeds, and costs. Choose the model that best fits your requirements.

​
Available Models

The model selector is located in the sidebar:
app.py
model_options = [
    'gpt-3.5-turbo',
    'gpt-4',
    'gpt-4-turbo',
    'gpt-4o-mini',
    'gpt-4o',
]
selected_model = st.selectbox(
    label='Informe o llm que deseja:',
    options=model_options
)

​
Model Comparison

GPT-3.5 Turbo

Best for: Fast responses, simple queries, high volumeFast and cost-effective option for straightforward questions

GPT-4

Best for: Complex analysis, detailed reasoningMost capable model with superior reasoning abilities

GPT-4 Turbo

Best for: Balance of speed and capabilityFaster than GPT-4 with similar performance

GPT-4o

Best for: Latest features and optimizationsNewest optimized model with enhanced capabilities

GPT-4o Mini

Best for: Cost-conscious applicationsSmaller, faster, cheaper variant of GPT-4o

​
Detailed Model Characteristics

​
GPT-3.5 Turbo

  • ⚡ Fastest response time
  • 💰 Lowest cost per request
  • 📊 Good for straightforward questions
  • 🔄 High throughput capability

​
GPT-4

  • 🧠 Superior reasoning
  • 🎯 Excellent accuracy
  • 📖 Better context understanding
  • 🔍 Complex query handling

​
GPT-4 Turbo

  • ⚡ Faster than GPT-4
  • 🧠 Near GPT-4 quality
  • 💰 Lower cost than GPT-4
  • 🎯 Great balance

​
GPT-4o (Optimized)

  • 🚀 Latest model
  • ⚡ Optimized performance
  • 🎯 Enhanced capabilities
  • 📊 Better efficiency

​
GPT-4o Mini

  • 💰 Very cost-effective
  • ⚡ Fast responses
  • 🎯 Better than GPT-3.5
  • 📊 Good balance for most tasks

​
Speed vs. Accuracy Tradeoffs

There’s typically a tradeoff between speed and quality. Faster models are generally less expensive but may produce less nuanced responses.

​
Cost Considerations

​
Relative Cost Comparison

ModelRelative CostSpeedQuality
GPT-3.5 Turbo$⚡⚡⚡⚡⚡⭐⭐⭐
GPT-4o Mini$$⚡⚡⚡⚡⭐⭐⭐⭐
GPT-4 Turbo$$$⚡⚡⚡⭐⭐⭐⭐⭐
GPT-4$$$$⚡⚡⭐⭐⭐⭐⭐
GPT-4o$$$⚡⚡⚡⭐⭐⭐⭐⭐
Actual costs vary based on OpenAI’s current pricing. Check OpenAI’s pricing page for current rates.

​
Recommendation Guide

​
Choose GPT-3.5 Turbo If:

  • You need fast responses
  • You’re working with simple questions
  • Cost is a primary concern
  • You have high query volume

​
Choose GPT-4o Mini If:

  • You want better quality than GPT-3.5
  • Cost efficiency is important
  • You need good general-purpose performance
  • You’re in development/testing phase

​
Choose GPT-4 Turbo If:

  • You need production-grade quality
  • You want a balanced option
  • Your queries are moderately complex
  • This is the recommended default for most users

​
Choose GPT-4 If:

  • You need maximum accuracy
  • You’re analyzing complex technical content
  • Quality is more important than speed
  • You’re working with critical applications

​
Choose GPT-4o If:

  • You want the latest model
  • You need optimized performance
  • You want cutting-edge capabilities
  • Speed and quality are both important

​
Changing Models

You can change the selected model at any time using the sidebar dropdown:
app.py
response = ask_question(
    model = selected_model,  # Uses the currently selected model
    query = question,
    vector_store = vector_store
)
Try different models with the same question to compare responses and find the best fit for your use case.

​
How Models Are Used

The selected model is passed to ChatOpenAI:
app.py
def ask_question(model, query, vector_store):
    llm = ChatOpenAI(model = model)
    retriever = vector_store.as_retriever()
    # ... rest of the function
This creates an instance of the LLM with your selected model for each question.

​
Next Steps

Asking Questions

Learn how to ask effective questions

Configuration

Configure API keys and other settings

Build docs developers (and LLMs) love

Get started for freeTalk to us