Windsurf Launches SWE-1 AI Models to Revolutionize Software Engineering

On June 5, 2025, Windsurf, a startup specializing in AI tools for software engineers, announced the launch of its first family of AI software engineering models called SWE-1. This launch marks a significant step as Windsurf moves beyond merely developing AI-powered applications to creating the foundational AI models themselves.

The SWE-1 family includes three models: SWE-1, SWE-1-lite, and SWE-1-mini. These models are designed to optimize the entire software engineering process, not just code generation. Windsurf emphasizes that software engineering involves complex workflows across multiple surfaces such as terminals, integrated development environments (IDEs), and the internet, which traditional AI models often struggle to handle effectively.

According to Windsurf, SWE-1 performs competitively with leading AI models like Claude 3.5 Sonnet, GPT-4.1, and Gemini 2.5 Pro on internal programming benchmarks, although it does not yet surpass frontier models such as Claude 3.7 Sonnet. The SWE-1-lite and SWE-1-mini models will be freely available to all users on Windsurf’s platform, while SWE-1 will be accessible exclusively to paid subscribers.

Windsurf’s Head of Research, Nicholas Moy, highlighted the distinction between coding and software engineering, stating that while frontier AI models have made significant progress in coding, they fall short in addressing the broader software engineering workflow. SWE-1 was trained using a novel data model and training recipe that incorporates incomplete states, long-running tasks, and multi-surface interactions to better reflect real engineering environments.

This launch is particularly notable given reports that OpenAI is in the process of acquiring Windsurf for $3 billion. Despite this, Windsurf appears committed to expanding its capabilities by developing proprietary AI models that power its applications, signaling a strategic move to differentiate itself in the competitive AI software engineering space.

Windsurf is best known for its “vibe coding” tools, which enable software engineers to write and edit code through conversational AI chatbots. This approach has gained popularity alongside other startups like Cursor and Lovable, which also rely on AI models from providers such as OpenAI, Anthropic, and Google. By developing SWE-1, Windsurf aims to offer a more integrated and efficient AI solution tailored specifically to the complexities of software engineering workflows.

The SWE-1 models represent Windsurf’s initial proof of concept in AI model development, suggesting that the company plans to continue innovating and releasing more advanced models in the future. This development underscores a broader industry trend where startups are not only leveraging existing AI models but are also investing in creating proprietary models to better serve niche use cases like software engineering.

By addressing the full spectrum of software engineering tasks—including managing incomplete states and long-running processes across various tools—Windsurf’s SWE-1 models have the potential to significantly enhance developer productivity and workflow integration. This advancement could lead to more seamless collaboration between AI and human engineers, ultimately accelerating software delivery and innovation.