Cerebras · OpenAI · Claude · Codex · GPT · Claude Code · cerebras.ai
Companies operating web services can leverage Ultrafast to root-cause and address production outages
Compiled by KHAO Editorial — aggregated from 2 sources. See llms.txt for citation guidance.
✓ KHAO Verified
Today, Cerebras and OpenAI are sharing an early look at Ultrafast Mode, a new service tier launching first in the OpenAI API and powered by Cerebras.
Key facts
- Benchmarking was performed by Cerebras using GPT 5.6 Sol Ultrafast with Codex on xhigh reasoning on July 10 and Claude Fable 5 with Claude Code on xhigh reasoning on July 13-15
- Benchmarking was performed by Cerebras on July 31 2026 using GPT 5.6 Sol and GPT 5.6 Sol Ultrafast on medium reasoning within Codex
- In their evaluations, GPT-5.6 Sol on Ultrafast mode answered all 2,500 HLE questions in 11 hours and 11 minutes
- Claude Fable 5 needed 78 hours and 27 minutes, more than three days of continuous compute, to arrive at the same conclusions
Summary
AI builders have always needed to choose between speed and intelligence. GPT-5.6 Sol Ultrafast resolves this tradeoff, bringing frontier intelligence to products and workflows where every second matters. At Cerebras, they put Ultrafast to the test by running it head-to-head with popular models on Humanity's Last Exam. In their evaluations, GPT-5.6 Sol on Ultrafast mode answered all 2,500 HLE questions in 11 hours and 11 minutes.