Skip to main content

​
Overview

Engines are the AI models that power Sphinx’s reasoning, code generation, and analytical capabilities. Each engine offers a different balance of speed, accuracy, and cost. Sphinx exposes a set of engines for data science workflows: code generation, data analysis, and visualization.

​
Available Engines

Sphinx offers two stable engines ready for production use. Additional beta engines are available in Engine Details.

Memphis

Default EngineSphinx’s frontier engine, ideal for complex coding and multi-step analytical tasks.
  • Primary Model: GPT-5.2
  • Context: 400K tokens

Thebes

Original EngineThe original Sphinx engine, a reliable general-purpose workhorse.
  • Primary Model: GPT-4.1
  • Context: 1M+ tokens

​
Engine Details

​
Memphis (Default)

Memphis is Sphinx’s default engine, powered by GPT-5.2. It handles complex coding and multi-step analytical tasks.
AttributeValue
ModelGPT-5.2
Context Window400,000 tokens
StatusStable
Memphis includes specialized behavior for data science workflows:
  • Works in small, concrete steps (one operation per cell)
  • Separates cleaning, feature engineering, and modeling
  • Avoids unnecessary reruns of successful code
Memphis is selected by default for new sessions. Use it when you’re unsure which engine to pick.

​
Thebes

Thebes is the original Sphinx engine—a proven, general-purpose model for everyday data science work.
AttributeValue
ModelGPT-4.1
Context Window1,047,576 tokens
StatusStable
Thebes is ideal when you need:
  • Maximum context window - great for your largest notebooks
  • Reliable, consistent performance

​
Siwa

Siwa is Sphinx’s speed and cost optimized engine, primarily powered by Claude Haiku 4.5. Use Siwa when you need fast iterations or are working on simpler tasks.
Siwa is still in beta and may not always behave as expected. Some features may be missing or have inconsistencies.
AttributeValue
ModelClaude Haiku 4.5
Context Window200,000 tokens
StatusBeta
Siwa is the right choice for:
  • Rapid prototyping and iteration
  • Simple data transformations
  • Cost-conscious workflows

​
Cairo

Cairo is Sphinx’s most powerful engine, built on Claude Sonnet 4.5. It delivers high quality reasoning for demanding analytical tasks.
Cairo is in early beta. Some features may not be fully polished.
AttributeValue
ModelClaude Sonnet 4.5
Context Window200,000 tokens
StatusBeta
Cairo excels at:
  • Nuanced code generation with thorough edge case handling
  • Detailed explanations and well-documented code
  • Designing polished, aesthetic visualizations and plots

​
Choosing the Right Engine

Choose Memphis for most analytical work, especially:
  • Multi-step data pipelines
  • Complex modeling tasks
  • When you want frontier capabilities without thinking about engine selection
Select Thebes when:
  • You need to work with extremely large context (1M+ tokens)
Use Siwa when:
  • Speed matters more than perfection
  • You’re doing rapid prototyping
  • The task is relatively simple
  • You want to minimize costs
Switch to Cairo when:
  • You want thorough edge case handling
  • You need presentation-ready visualizations
  • Quality matters more than speed
  • You prefer Claude’s coding style and reasoning over GPT’s

​
Switching Engines

You can switch engines at any time during your session using the engine selector in the Sphinx interface. The change takes effect immediately for subsequent requests.
Switching engines mid-session does not affect code that has already been generated or executed. The new engine will be used for all future requests in that session.