OpenAI Launches 'Ultrafast' With GPT-5.6 Sol Running 14x Faster
The Spin
Pro-establishment narrative
Ultrafast mode is a genuine leap forward — GPT-5.6 Sol running at 750 tokens per second means frontier intelligence finally fits into real-time, high-stakes work like incident response, financial analysis and live customer support. Clearing Humanity's Last Exam in just over 11 hours proves this isn't a gimmick. Speed at this level, without sacrificing intelligence, fundamentally changes what AI can do for businesses.
Establishment-critical narrative
Ultrafast mode is really just a faster way to drain your compute budget — at 14x speed, billing meters spin 14x faster, and parallel agent workflows will eat through context windows and weekly limits before most users know what hit them. The monetization logic is almost cynical — better agents create more work, faster inference burns through it, and paid resets cash in on the fallout. This is a revenue flywheel dressed up as a feature.