SWE-2

Name: SWE-2
Author: Cognition

Coding model

SWE-2

Release date: 2026-09-10Views: 0

Live demoGitHubHugging FaceCompare

Parameters

No data

Context length

No data

Multilingual

No data

Reasoning ability

No data

SWE-2 is an AI model published by Cognition, released on 2026-09-10, for Coding model, with a 92.80 score on Terminal-Bench 2.1.

Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology

SWE-2

Model basics

Reasoning traces

Supported

Thinking modes

Thinking Level · High (Default)Thinking Level · MediumThinking Level · Max

Context length

No data

Max output length

No data

Model type

Coding model

Modality (in / out)

Text → Text

Release date

2026-09-10

Model file size

No data

MoE architecture

Total params / Active params

No data / Not applicable

Knowledge cutoff

No data

SWE-2

Open source & experience

Code license

Proprietary

Weights license

Proprietary

GitHub repo

N/A

Hugging Face

N/A

Live demo

devin.ai/desktop

SWE-2

Official resources

Paper

DataLearnerAI blog

N/A

SWE-2

API details

API speed

No data

No public API pricing yet.

SWE-2

Benchmark Results

SWE-2 currently shows benchmark results led by Terminal-Bench 2.1 (1 / 193, score 92.80), Terminal-Bench 4.0 (20 / 87, score 27.30), FrontierCode 1.1 Main (4 / 5, score 50). This page also consolidates core specs, context limits, and API pricing so you can evaluate the model from benchmark results and deployment constraints together.