Devin users get a strong coding model that is free on self-serve plans only until mid-October, then billed per token.
Kenning's take
What it does
SWE-2 is Cognition's in-house model for software engineering, used by the Devin coding agents. It comes in Medium, High and Max reasoning-effort versions, and Cognition says it scores close to leading frontier models on coding benchmarks at a lower cost. It is available in Devin Desktop and the Devin CLI, and is rolling out to Devin Web and Fusion.
- Medium, High and Max reasoning effort
- Post-trained from Kimi K3 with reinforcement learning
- 50.0% on FrontierCode 1.1 Main, per Cognition
- Fewer steps per task than SWE-1.7
- Available in Devin Desktop and CLI
Who it suits. Developers using Devin Desktop, the Devin CLI or Devin Web for coding agents.
Key facts
- Category
- AI models
- Made by
- Cognition
- Released
- 10 September 2026
- How to get it
- in Devin Desktop and Devin CLI, rolling out to Devin Web and Fusion; free on Devin Pro for a limited period
- Works on
- Devin Desktop, Devin CLI, Devin Web
- Free plan
- Devin's free plan has limited model availability; SWE-2 is free on self-serve plans through 15 October 2026 per the docs (the pricing page says 16 October)
- API price
- $3.00 per million input tokens, $15.00 per million output, $0.30 per million cached input (list price after the free period)
Pricing
| Plan | Price | Billed |
|---|---|---|
| Pro | $20 | month |
| Max | $200 | month |
| Teams | $80 plus 40 per full dev seat | month |
FAQ
When was SWE-2 released?
Cognition introduced SWE-2 on 10 September 2026.
Is SWE-2 free?
It is free on self-serve Devin plans until mid-October 2026. After that, list pricing is $3 per million input tokens and $15 per million output.
Where can I use SWE-2?
In Devin Desktop and the Devin CLI, with Devin Web and Fusion rolling out.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.