SWE-2

AI model · Code model · Text → Code by Codeium

High-performance, cost-efficient autonomous coding model.

SWE-2 is Cognition's most advanced proprietary coding model, designed to provide frontier-level performance for autonomous software engineering at a fraction of the cost. It is integrated into the Devin Desktop and CLI environments and powers free concurrent sessions for Teams users.

Model specs

The full spec card (context, pricing per token, modalities and more) on AntSpace. Sign up free to see the full spec card.

Launched
2026-11-12
Languages
English

Category: Generative AI

Modalities
Text (code) to text (code)
Editors and tools
Devin Desktop, Devin CLI
License
Proprietary
Access
Devin Desktop, Devin CLI, Devin Cloud (Free mode)
Released
2026-11-12
Type
Coding Model
  • Cognition's most advanced coding model to date
  • Performs close to frontier models at a fraction of the cost
  • Powers Free mode sessions for Devin Teams users
  • Permanent part of Teams, Pro, and Max plans for Desktop and CLI

Overview

SWE-2 is a proprietary coding model developed by Cognition, designed for autonomous software engineering tasks within the Devin ecosystem. It serves as Cognition's most advanced model to date, offering performance levels close to frontier models while maintaining a significantly lower operational cost. The model is optimized for speed and efficiency, specifically focusing on the relevant sections of a codebase to accelerate task completion. SWE-2 is integrated into Devin Desktop and the Devin CLI, and it powers the Free mode sessions for Teams users, allowing for concurrent agent usage. It is available as a standard feature for users on the Teams, Pro, and Max plans.

Core Capabilities

  • Advanced Coding Performance: SWE-2 is Cognition's most advanced coding model to date, designed for high-performance software engineering tasks.
  • Frontier-Level Efficiency: The model performs close to frontier models while operating at a fraction of their cost.
  • Efficient Codebase Focus: SWE-2 focuses on the specific parts of a codebase that matter for a given task to start working faster.

Access and Integration

  • Free Mode Power: SWE-2 powers the Free mode sessions for Devin Teams users, allowing for concurrent agent usage.
  • Desktop and CLI Integration: The model is a permanent part of the Devin Desktop and CLI environments for Teams, Pro, and Max users.
  • Quota Management: Users can switch to SWE-2 for routine tasks to extend their usage quotas on paid plans.

Use cases

  • Software engineers use SWE-2 in Devin Desktop to build features and fix bugs with frontier-level performance.
  • Teams users run concurrent Free mode sessions powered by SWE-2 to handle multiple coding tasks simultaneously.
  • Developers switch to SWE-2 for routine coding tasks to extend their usage quotas on paid Devin plans.

Specifications

Modalities
Text (code) to text (code)
Editors and tools
Devin Desktop, Devin CLI
License
Proprietary
Access
Devin Desktop, Devin CLI, Devin Cloud (Free mode)
Released
2026-11-12
Type
Coding Model
Developer
Cognition
Access Methods
Devin Desktop, Devin CLI, Devin Cloud
Concurrent Sessions (Teams)
2 per user in Free mode

How it works

  1. Select a Plan: Users sign up for a Devin plan such as Teams, Pro, or Max to gain access to the SWE-2 model.
  2. Access via Desktop or CLI: SWE-2 is integrated into the Devin Desktop and CLI environments, where it can be selected for coding tasks.
  3. Run Free Mode Sessions: For Teams users, SWE-2 can be used in Free mode to run up to 2 concurrent sessions without consuming standard quotas.

Pricing

Free use of SWE-2 (Pro Plan)
$20 per month
Free use of SWE-2 (Max Plan)
$200 per month
2 concurrent Free mode sessions per user (Teams Plan)
$80 per month

FAQ

What is SWE-2?

SWE-2 is Cognition's most advanced coding model that performs close to frontier models at a fraction of their cost.

How can I access SWE-2?

SWE-2 is available as a permanent part of the Teams, Pro, and Max plans for use in Devin Desktop and CLI. It also powers the Free mode sessions for Teams users.

How does SWE-2 handle large codebases?

SWE-2 is designed to get to work faster by focusing specifically on the parts of the codebase that are relevant to the task at hand.

Is there a free version of SWE-2?

Yes, Teams users get 2 concurrent Free sessions powered by SWE-2 as part of their plan.

Get started

Added by AntSpace from public sources (devin.ai).

Related

More AI models by Codeium