Skip to main content
Smart model routing is Superblocks’ cost-optimization mode for AI app building. When enabled, smart model routing automatically chooses the right model for the right task, balancing frontier performance with lower-cost model execution. Instead of sending every step of a session to the most expensive model, Superblocks intelligently routes work across frontier and open-source models to reduce inference costs while preserving output quality. Customers can expect approximately 20-25% cost savings per session, depending on the task mix.

​
Why smart model routing

Frontier Lab products like Claude Code or Codex use frontier models only and do not offer cheaper Open Source models, eliminating their ability to achieve the best price performance. As the landscape changes, the best model for one task may not be the best model for another. Some tasks require maximum frontier intelligence. Others are simpler, repetitive, or implementation-oriented and can be completed by more cost-effective models. Furthermore, AI model releases alongside price and performance are changing at a pace too fast for most Superblocks Admins or Superblocks Builders to keep up. Smart model routing lets Superblocks optimize price and performance automatically on every prompt. Customers do not need to manually select models, benchmark providers, or manage routing logic. They choose the experience they want, and Superblocks handles the model orchestration behind the scenes.

​
Available modes

Superblocks offers three AI execution modes: Max, Standard, and Smart.
  • Max: The max intelligence for frontier performance on the task.
  • Standard: Frontier intelligence appropriate for most coding tasks.
  • Smart: Best performance at lowest cost. Superblocks decides when to use frontier or open source to cost-optimize your task.

​
How smart model routing works

When a builder starts a task, Superblocks evaluates the request and routes work to the right model. Complex planning, reasoning, and code-generation steps may use frontier models. Simpler edits, transformations, summaries, or routine implementation steps may use cost-effective models. Superblocks continuously evaluates quality, latency, and cost. As model performance changes and new models launch, Superblocks updates routing decisions without requiring customers to reconfigure their workspace. This lets customers benefit from new models as they become available without managing model complexity themselves.