CognitionLanguage modelAvailable in Devin
SWE-2 Flash
Summary
SWE-2 Flash is Cognition's model for routine software tasks in Devin, such as reviewing code and making small project changes. Cognition trained the Inkling-Small base model further for these tasks.
The model is available in Devin Desktop and the Devin CLI. Cognition reports 276 billion total parameters, with 12 billion active.
Similar models
More models of the same kind as SWE-2 Flash, chosen by model line, developer, size, price and release date.
- ERNIE 5.5 Flash PreviewBaidu · similar size
- SWE-2Cognition · same developer
- GLM-5.3-FlashZ.ai · similar size
- DeepSeek V4 Flash 0731DeepSeek · similar size
- Inkling SmallThinking Machines Lab · similar size
- DeepSeek-V4-FlashDeepSeek · similar size
- Hy3Tencent · similar size
- ERNIE 5.5 Flash BaseBaidu · similar size
All specifications
Architecture
- Base model
- Cognition trained Thinking Machines Lab's Inkling-Small further for software tasks.
- Source: Base model
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
- Shutdown
- None announced
- no source
Prices
- Launch offer
- Cognition advertises free usage in Devin through November 30, 2026.Source
Frequently asked questions about SWE-2 Flash
Can I run SWE-2 Flash locally?
No. Cognition does not publish the weights.
Stated by Cognition
- Open weights
- no
- License
- n/a
- Parameters
- 276B
- active per token
- 12B
Sources and data date
| Type | Evidence and data date |
|---|---|
| Additional source | Cognition: Introducing SWE-2 Flash (retrieved November 12, 2026) · Cognition: Introducing SWE-2 Flash · Editorial description reviewed November 12, 2026 |