OpenAI speeds up Ultrafast mode for GPT‑5.6 Sol by 14 times

Summary

OpenAI introduced the Ultrafast mode for GPT‑5.6 Sol; leveraging Cerebras hardware to boost inference speed by up to 14× (750 tokens per second), targeting time-sensitive workflows.

Why it matters

This speed boost eliminates the traditional trade‑off between model size and latency, enabling high‑intelligence AI to be integrated into real‑time applications such as incident response, financial monitoring, and customer support. It could shift AI from batch processing to interactive, enterprise‑level decision‑making.

Highlights

• 14× speed increase without compromising GPT‑5.6 Sol's intelligence.
• Powered by Cerebras, delivering 750 output tokens per second.
• Early adopters report transformative impacts in voice AI, finance, and commerce.

Critical take

The preview is limited to a select customer group, and reliance on specialized hardware like Cerebras could create cost and accessibility barriers for smaller businesses.