OpenAI's GPT-5.6 Sol (max) earns the highest Presentation Elo in the AA-Briefcase benchmark, completing tasks 61% faster at half the cost of competitors.

OpenAI's GPT-5.6 Sol targets 750 tokens per second via Cerebras wafer-scale compute, potentially outperforming GPU clusters by 15x in AI inference tasks.

OpenAI dropped GPT-5.6 in three tiers. But the real problem in AI coding has not changed.