Cerebras powers OpenAI GPT-5.6 Sol Ultrafast mode at 14x speed

Cerebras Systems (NASDAQ: CBRS) announced it is powering Ultrafast mode, a new service tier in the OpenAI API for GPT-5.6 Sol. The service is available initially in limited preview to OpenAI customers. Andrew Feldman, CEO and co-founder of Cerebras, stated that the partnership demonstrates that speed and intelligence are no longer mutually exclusive.

Ultrafast runs GPT-5.6 Sol at up to 750 output tokens per second. This represents processing speeds of up to 14 times faster than Standard processing. Sachin Katti, VP Compute Strategy & GPT-Infra at OpenAI, noted that the company is starting with a small group of customers to learn where that speed creates meaningful value before expanding the service.

Performance Benchmarks

The performance differential highlights the infrastructure advantage provided by Cerebras. While both modes deliver the same intelligence as GPT-5.6 Sol Standard, the Ultrafast tier achieves this frontier intelligence at significantly higher throughput. Based on output speeds for Anthropic models reported by Artificial Analysis, Ultrafast is 5x faster than Claude Opus 4.8 in Fast mode and 11x faster than Claude Fable 5.

Benchmarks focused on economically valuable work show how faster token generation translates into higher productivity:

On Humanity’s Last Exam, a 2,500-question benchmark spanning graduate-level chemistry, economics and literature, GPT-5.6 Sol Ultrafast answered the full question set in just over 11 hours. This compares to more than three days of continuous compute for Claude Fable 5, with GPT-5.6 Sol Ultrafast reaching comparable accuracy nearly 7x faster.
On GDP-Val, a benchmark of economically valuable knowledge-work tasks such as legal briefs, financial models, and engineering reports, Ultrafast delivered a 5.6x end-to-end speedup with no loss in quality.

What the Numbers Show

The data reveals a divergence between raw model capability and inference efficiency. By maintaining the same intelligence level as the Standard tier while achieving a 14x speed increase, Cerebras’ architecture effectively decouples latency from model size. The comparison with Claude models further contextualizes this advantage: being 5x faster than Claude Opus 4.8 suggests that Cerebras’ Wafer-Scale Engine architecture provides a distinct competitive edge in high-throughput inference scenarios where memory bandwidth is typically the bottleneck.

Technical Architecture

Ultrafast’s speed comes from Cerebras’ Wafer-Scale Engine architecture, which keeps model weights on-chip — 44 GB of SRAM on each wafer-sized chip — rather than shuttling them between on-chip memory and off-chip storage as GPU-based inference must. This eliminates the memory-bandwidth bottleneck that constrains frontier-model inference speed on conventional hardware.

To get notified when capacity expands to more customers, please visit the Cerebras website.