by Cognition
SWE-2 is Cognition's coding model for agentic software engineering, built to deliver strong results at a lower cost per task. It is available in Devin Desktop and the Devin CLI.
APIContact for pricing
Step 1: Open Devin Desktop or the Devin CLI, where SWE-2 is available.
Step 2: Select SWE-2 as the model for your session.
Step 3: Pick an effort level: medium for simple and intermediate tasks, high or max for complex ones.
Step 4: Describe the coding task and let the agent explore the codebase, edit and run tests.
Step 5: Review the changes and the verification the agent ran before accepting them.
Cognition says SWE-2 is available in Devin Desktop and the Devin CLI, and is being rolled out on Devin Web.
The post describes medium, high and max. Medium acts quicker on simple and intermediate tasks, while high and max plan and verify more on complex ones.
Cognition, the applied AI lab that also operates Devin.
E2BE2B gives every AI agent its own isolated cloud machine, a microVM sandbox where it can run code, browse and use tools. It runs in E2B's cloud or in your own AWS, GCP or Azure account.
Meta Model APIMeta Model API is a self-serve inference API for Meta's Muse models, covering text and agentic work, image generation, speech transcription and image segmentation. It works with existing OpenAI SDK code.