Skip to main content

Adaptive

For most users, we recommend Adaptive — our intelligent model router that automatically selects the best model for each task, delivering the right level of intelligence for every prompt.
Some models are Devin Local only, including every GPT-5.6 variant. They appear disabled in Cascade’s model picker with a tooltip pointing you to the Devin Local agent — switch agents to use them.
You can easily switch between different models of your choosing. Under the text input box, you will see a model selection dropdown menu containing the following models:
For the most up-to-date pricing and availability, please refer to the model selector in the agent panel.
Your quota and extra usage is billed based on the token cost of the model you select. You can view the cost of each model in the table below.

​
SWE-1.7, swe-grep, swe-check

Our SWE model family of in-house frontier models are built specifically for software engineering tasks. Our latest model, SWE-1.7, is available as a free preview until August 8. SWE-1.7 Lightning runs on Cerebras for an even faster experience. Our in-house models include:
  • SWE-1.7: Cognition’s latest software engineering model, available free during the preview.
  • SWE-1.7 Lightning: A faster version of SWE-1.7 served on Cerebras, delivering the same intelligence with lower latency.
  • SWE-1.6: Our previous-generation model built for software engineering agents, optimized for both intelligence and model UX. Read our research announcement.
  • SWE-1.6 Fast: A faster version of SWE-1.6 available to paying users.
  • SWE-1-mini: Powers passive suggestions in Tab, optimized for real-time latency.
  • swe-grep: Powers context retrieval and Fast Context
  • swe-check: Powers Quick Review with fast, lightweight reviews optimized for common code issues.