OpenAI’s two new models reached the top 10 of Code Arena’s WebDev leaderboard, which relies on crowdsourced community evaluations. GPT-6.2 Astra (Max) scored 1,781 points to take third place, while GPT-6.2 Sol (Max) scored 1,757 points to finish seventh. Both models reached the performance-cost Pareto frontier, with median cost per task falling by 42% for Astra and 21% for Sol. Astra leads Sol by 24 points and costs about 3.8 times more per task, but it took first place in the Consumer Product category. The models launched on November 17, 2026, and their agentic task scores remain pending.
Source: Read the original article


