OpenAI and Cerebras Launch Ultrafast Mode for GPT-5.6 Sol

Aug 13, 2026

AI Summary

OpenAI and Cerebras have introduced Ultrafast Mode for the GPT-5.6 Sol model, enabling output speeds of up to 750 tokens per second without compromising quality. This new service aims to enhance efficiency for time-sensitive tasks across various industries, significantly outperforming previous models in speed and accuracy.

Ultrafast Mode is a new service tier for the OpenAI API, powered by Cerebras, initially available to select customers.
GPT-5.6 Sol in Ultrafast Mode can deliver up to 750 output tokens per second, addressing the trade-off between speed and intelligence in AI models.
The model runs 11 times faster than Fable 5 and 5 times faster than Opus 4.8, showcasing significant performance improvements.
In a benchmark test, GPT-5.6 Sol answered 2,500 challenging questions in just over 11 hours, while competitors took significantly longer.
Ultrafast Mode has shown a 5.6 times speedup on economically valuable tasks without quality degradation, making it suitable for legal, financial, and engineering applications.
The technology is based on Cerebras’ Wafer-Scale Engine architecture, designed to minimize data movement and improve processing efficiency.
Ultrafast Mode is expected to expand access as capacity increases, providing organizations with rapid AI responses for critical operations.

gpt-5openaiultrafastai researchlanguage models

OpenAI and Cerebras Launch Ultrafast Mode for GPT-5.6 Sol

Related Stories

Advice on Learning to Build Language Models at a Young Age

Improvements in LLM-Assisted Coding Through Customization and Agentic IDEs

Local LLM Performance Affected by Hardware and Software Variations

Anthropic conducts A/B testing on effort levels in Claude Code