Cognition releases SWE‑2 Flash for faster, lower-cost coding work
The 276-billion-parameter SWE‑2 Flash is available in Devin Desktop and CLI at one-fifth the price and three times the speed of SWE‑2.
AT A GLANCE
Key points
- SWE‑2 Flash has 276 billion total parameters and 12 billion active parameters.
- Cognition says it runs three times faster than SWE‑2 at one-fifth the price.
- It is available in Devin Desktop and CLI and is free through November 30.
What is changing?
Cognition announced its SWE‑2 Flash coding model on November 4. The company says the model has 276 billion total parameters and 12 billion active parameters, and was developed from Thinking Machines’ Inkling‑Small base through mid-training and reinforcement learning. It is available today in Devin Desktop and CLI.
Cognition says SWE‑2 Flash runs three times faster than its larger SWE‑2 sibling at one-fifth the price, reaching 200 tokens per second in the company’s example. Its reported scores include 24.0 on FrontierCode 2.0 and 20.0 on Terminal‑Bench 4.0. These are provider measurements, not independent verification.
The model is free through November 30. Its weights are not public; citing cyber risks, Cognition says it is building a controlled sharing program for trusted industry partners and research labs.
Explore the original sources
Source date: 4 November 2026.
Published by AI Radar: 5 Nov 2026, 09:08 (Istanbul, UTC+3).
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…