SWE-2 is Cognition’s coding model, post-trained from Kimi K3 with reinforcement learning.

Where is SWE-2 available?

It is available in Devin Desktop and CLI.

What benchmark result is reported for SWE-2?

The description reports a 50.0% result on FrontierCode 1.1 Main.

How does SWE-2 compare with SWE-1.7?

The description reports 58% fewer turns, 81% lower cost, and a higher score than SWE-1.7.

What cost comparisons are stated for SWE-2?

The description states that SWE-2 is 64% less expensive than Fable 5.1 and costs one quarter as much as GPT-6 Astra.

Coding / product dossier

Cognition's SWE-2

A coding model in Devin Desktop and CLI optimized for lower cost and fewer turns.

Open product Daily selection

Product brief

What Cognition's SWE-2 does.

SWE-2 is Cognition's new coding model, post-trained from Kimi K3 with RL that optimizes for cost and capability at the same time. It hits 50.0% on FrontierCode 1.1 Main, within a point of Fable 5.1 at 64% less, and lands within a few points of GPT-6 Astra at a quarter of the cost. Compared to SWE-1.7 it takes 58% fewer turns and costs 81% less while scoring higher. Available now in Devin Desktop and CLI.

Superpower Daily

Your reading space

Notifications

Superpower Daily

Your reading space

Cognition's SWE-2

What Cognition's SWE-2 does.

What it can help with.

Support coding tasks

Use post-training optimized for cost and capability

Operate in Devin workflows

Reduce turns versus SWE-1.7

Before you open it.

When we picked it.

Cognition's SWE-2

What Cognition's SWE-2 does.

What it can help with.

Support coding tasks

Use post-training optimized for cost and capability

Operate in Devin workflows

Reduce turns versus SWE-1.7

Before you open it.

When we picked it.

Codync

ZooWork

JarvisCore