Windsurf Launches New AI Models for Software Engineers
Windsurf, a startup that develops popular AI tools for software engineers, has announced the launch of its first family of AI software engineering models, known as SWE-1. This new family of models, which includes SWE-1, SWE-1-lite, and SWE-1-mini, has been trained to optimize the entire software engineering process, not just coding.
What Makes SWE-1 Special
The launch of Windsurf’s in-house AI models may come as a surprise, given that OpenAI has reportedly closed a $3 billion deal to acquire Windsurf. However, this move suggests that Windsurf is trying to expand beyond just developing applications to also developing the models that power them. According to Windsurf, SWE-1 performs competitively with other top AI models, such as Claude 3.5 Sonnet, GPT-4.1, and Gemini 2.5 Pro, on internal programming benchmarks.
Comparison with Other AI Models
Although SWE-1 appears to fall short of frontier AI models, such as Claude 3.7 Sonnet, on software engineering tasks, it still offers impressive capabilities. Windsurf claims that SWE-1 is cheaper to serve than Claude 3.5 Sonnet, making it a more affordable option for software engineers.
Availability and Pricing
Windsurf’s SWE-1-lite and SWE-1-mini models will be available for all users on its platform, free or paid. Meanwhile, SWE-1 will only be available to paid users. Although Windsurf did not announce pricing for its SWE-1 models, it promises to provide more information in the future.
Windsurf’s Approach to AI
Windsurf is best known for its tools that allow software engineers to write and edit code through conversations with an AI chatbot, a practice known as "vibe coding." The company’s Head of Research, Nicholas Moy, notes that while today’s frontier models are optimized for coding, they are not enough for software engineering. Moy emphasizes that coding is not the same as software engineering, and that Windsurf’s new models are designed to address this gap.
Training SWE-1
Windsurf trained SWE-1 using a new data model and a "training recipe that encapsulates incomplete states, long-running tasks, and multiple surfaces." This approach allows SWE-1 to work seamlessly between multiple surfaces, such as terminals, IDEs, and the internet, making it a more versatile tool for software engineers.
Future Developments
Windsurf describes SWE-1 as its "initial proof of concept," suggesting that it may release more AI models in the future. This move marks an exciting development in the field of AI software engineering, and it will be interesting to see how Windsurf’s new models impact the industry.
Conclusion
The launch of Windsurf’s SWE-1 models is a significant step forward for the company and the field of AI software engineering. With its focus on optimizing the entire software engineering process, rather than just coding, Windsurf is poised to make a major impact on the industry. As the company continues to develop and refine its models, it will be exciting to see how they shape the future of software engineering.
