Accelerating GPT-5.6 Sol Ultrafast

	Accelerating GPT-5.6 Sol Ultrafast(cerebras.ai)
	707 points by pr337h4m 8 days ago \| 276 comments
	tl;dr: Cerebras and OpenAI have launched "Ultrafast Mode" for GPT-5.6 Sol, delivering up to 750 output tokens/second—reportedly 11x faster than Fable 5 and completing Humanity's Last Exam in 11 hours versus 78 for Claude Fable 5. The speedup is enabled by Cerebras' Wafer-Scale Engine, which packs 44GB of SRAM per chip to keep model weights on-chip and eliminate the memory-bandwidth bottleneck that slows GPU inference. It's currently available as a limited preview to select OpenAI API customers.
	HN Discussion: ↑Excitement about speed improvements and their importance for iteration and quality of thought ↓Skepticism that Ultrafast mode maintains identical quality to regular GPT-5.6 Sol due to vague messaging ↓The comparison omits competing fast models like Mimo v2.5-Pro Ultraspeed, weakening the claims ~Faster token throughput doesn't eliminate other bottlenecks like tests, typechecks, and grep •Anticipation for specialized ASIC hardware enabling local, offline, ultra-fast inference