Cognition launches SWE-2 Flash, a faster, cheaper coding model for Devin

Cognition released SWE-2 Flash, a coding model built on Thinking Machines' Inkling-Small that it says runs at about three times the speed and one-fifth the price of its larger SWE-2 model, available in Devin Desktop and Devin CLI.

What happened

• SWE-2 Flash is built on Inkling-Small (276B total, 12B active parameters), with a new mid-training stage before reinforcement learning.
• Cognition reports its score on FrontierCode 2.0 rose from 3.2% for the base model to 24.0%, and on Terminal-Bench 4.0 from 0.9% to 20.0%.
• It's aimed at routine agent work such as automations, code review, feature development and documentation updates, and comes in Medium, High and Max effort variants.
• Devin's docs say it's free for self-serve plans until December 4, 2026, then $0.10 per million input tokens and $0.50 per million output tokens; free plans get the Medium variant.
• The weights aren't public; citing cybersecurity risk, Cognition says it is building a program to share them with vetted partners and researchers.

Sources

• Cognition: Introducing SWE-2 Flash cognition.com
• Devin docs: Chat models and pricing docs.devin.ai

Checked November 9, 2026 against the sources above. Claims about performance or benchmarks are the company's own unless stated otherwise.