SWE-1.7 free preview lands in Devin Desktop
Devin Desktop adds SWE-1.7, SWE-1.7 Lightning, Claude, and GPT options, with a free SWE-1.7 preview through Aug. 8.

Choosing a coding model often means trading speed for cost, then checking whether the task needs a heavier model at all. Devin Docs now lays out those options for Devin Desktop users, including the SWE family, Claude, and GPT.
Devin Desktop now lists model choices for coding tasks, including a free SWE-1.7 preview.
| 項目 | 數值 |
|---|---|
| SWE-1.7 preview end date | August 8 |
| SWE-1.7 Lightning host | Cerebras |
| Model families listed | SWE, Claude, GPT |
| Legacy enterprise billing | credits |
What changed
Get the latest AI news in your inbox
Weekly picks of model releases, tools, and deep dives — no spam, unsubscribe anytime.
No spam. Unsubscribe at any time.
The docs page now groups available models for Devin Desktop and Devin CLI, and it points users to a model selector in the agent panel for current pricing and availability. It also notes that some models are Devin Local only, including every GPT-5.6 variant, which are disabled in Cascade’s model picker until users switch agents.

The in-house SWE lineup is the main addition. SWE-1.7 is described as Cognition’s latest software engineering model and is free during preview until August 8, while SWE-1.7 Lightning runs on Cerebras for lower latency. The page also lists SWE-1.6, SWE-1.6 Fast, SWE-1-mini, swe-grep, and swe-check.
- SWE-1.7: latest software engineering model, free in preview
- SWE-1.7 Lightning: same model class, faster response time
- SWE-1-mini: powers passive suggestions in Tab
- swe-grep: used for context retrieval and Fast Context
- swe-check: powers Quick Review for common code issues
Why it matters
For developers, the practical change is less about a brand-new product than about clearer model routing. The docs now spell out which model fits which job, so teams can choose a lighter option for autocomplete, a faster one for quick feedback, or a stronger model for larger engineering tasks.

For enterprise customers, the billing split matters too. Legacy plans still track usage by token cost and credits, while newer enterprise plans are billed in ACUs, which means model choice can affect both speed and spend. That makes the selector more than a convenience feature.
The bigger question is whether users will stick with Adaptive, the router that picks a model automatically, or keep tuning model choice by hand when latency and cost start to matter.
// Related Articles
- [TOOLS]
Windsurf’s IntelliJ plugin is a shortcut, not a strategy
- [TOOLS]
Vibe Island’s changelog shows the right product bets
- [TOOLS]
OpenAI Newsroom turns announcements into a digest
- [TOOLS]
Trendshift monthly repos let you spot real momentum
- [TOOLS]
Rust vs Go in 2026: pick the right fit
- [TOOLS]
12 Cursor alternatives that cost less in 2026