- Who it helps
- Developers comparing coding quality, latency and access costs.
- Availability
- Launched November 4. The page advertises free use through November 30; public weights are not available.
What changed
Cognition added mid-training and reinforcement learning to Inkling-Small, a model with 276 billion total and twelve billion active parameters. Its own evaluations report gains over that base.
Price and speed need task-level checks
Cognition advertises one-fifth of SWE-2’s price and three times its generation speed. The page’s side-by-side animation replays shortened transcripts at simulated speeds; it is not an independent timing test. A weight-sharing program is planned for selected partners.
What you can do
- Read the launch conditions and compare a fixed task before choosing a default. Open reference ↗
Where the evidence stops
No coding task or billing check was executed. Relative speed does not establish faster completion of every task.
Sources read. Sources read November 5, 2026 UTC. No third-party code, model inference, account workflow or benchmark executed.
Sources & further reading
Web · official
SWE-2 Flash adds a faster coding option in Devin ↗
Identifies the subject, documented capability or specific verification limit.
Cognition · page checked 2026-11-05
Coverage note: Independent attention and end-to-end execution remain unverified.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
- · Added to Current
SWE-2 Flash is Cognition’s smaller coding model for Devin Desktop and CLI.