LlamaRestTest demonstrates how small, fine-tuned language models can effectively test REST APIs by learning to generate requests that find dependencies and errors.

• Achieves superior testing coverage compared to traditional tools and larger GPT models
• Requires only 1GB of RAM while matching or exceeding GPT-4's capabilities
• Discovers parameter dependencies and vulnerabilities missed by conventional testing approaches
• Fine-tuned on just ~125K examples to specialize in API testing

This innovation matters for engineering teams because it makes advanced API testing accessible without costly compute resources, ensuring more robust and secure web services.

LlamaRestTest: Effective REST API Testing with Small Language Models