We made Grok 4.5, GPT-5.5, and Claude build the same apps (www.tryai.dev)

xAI has launched Grok 4.5, its latest and most advanced model, designed for coding and agentic tasks, and it was put to the test against other leading models: GPT-5.5, Claude Opus 4.8, and Claude Fable 5. All models were tasked with creating three interactive apps using the same prompts. The results showcased their strengths and weaknesses, with Claude's models excelled in producing a functioning 3D Rubik's Cube on the first try, while Grok had a mixed performance but impressed with speed and cost efficiency. GPT-5.5 stood out for its visual appeal in the gravity sandbox, despite struggling with the cube task. The significance of this test lies in Grok 4.5's competitive edge in processing speed and operational costs, streaming at approximately 110 tokens per second and generating responses more affordably than its counterparts. This positions Grok 4.5 as an appealing option for high-volume coding tasks that prioritize rapid output and low expenses. Meanwhile, Claude's models offered reliability and creativity, particularly with Fable 5's entertaining SVG creations. Ultimately, the build-off highlights the varying capabilities of these AI models, while suggesting Grok 4.5's strong potential in practical applications within the AI/ML community.