OpenAI's GPT-5.6 Models Ranked on Fullstack Code Arena Benchmark
Arena.ai has published the Fullstack Code Arena benchmark scores for OpenAI's GPT-5.6 models, including Sol, Terra, and Luna. In the rankings, the flagship Sol model secured the #3 spot in the Fullstack category and #5 overall. As AI coding assistants evolve, benchmarks like Fullstack Code Arena provide critical insights into how models perform when building complete, production-grade web applications rather than just writing isolated code snippets. This helps developers choose the most capable and cost-effective models for real-world software development. The benchmark ranked the "xHigh" configurations of the models, placing Terra at #10 in Fullstack (#20 overall) and Luna at #14 in Fullstack. These rankings follow OpenAI's July 2026 release of the GPT-5.6 family and subsequent price reductions for Luna and Terra.
## BACKGROUND
Fullstack Code Arena, developed by Arena.ai (formerly LMArena), evaluates AI models on their ability to autonomously build deployable web applications with persistent databases. OpenAI launched the GPT-5.6 family in July 2026, featuring Sol as the flagship model, Terra as a balanced option, and Luna as a cost-efficient model.