Up to 14x Faster! Cerebras (CBRS.US) creates “ultra-high-speed mode” for OpenAI’s GPT-5.6 Sol model

According to Zhitong Finance APP, artificial intelligence (AI) chip rising star Cerebras Systems (CBRS.US) announced on Thursday that the company is providing a brand-new Ultrafast mode for OpenAI's GPT-5.6 Sol model. Reportedly, the new Ultrafast mode processes at speeds up to 14 times faster than OpenAI's standard mode and is currently being gradually rolled out to a small group of customers with higher demands for lower latency.

Cerebras stated on its website: "Ultrafast mode is initially available only to a small group of customers and will be gradually expanded in the future. Cerebras supports GPT-5.6 Sol’s Ultrafast mode with the computing power needed to deliver speeds of up to 750 output tokens per second without sacrificing any quality, enabling Sol Ultrafast to accelerate processing of the most time-sensitive and mission-critical workloads."

In another article, OpenAI said: "Their early usage is helping us understand where an order of magnitude increase in speed can create the most value, and how the product changes when the model can keep up with the user's pace. As computing capacity continues to expand, we'll use these insights to guide deployment." OpenAI also indicated that the company has already identified some "encouraging" applications for Ultrafast mode, including event response and reliability, financial research and security, customer support and voice, business, as well as real-time research and experimentation.

GPT-5.6 Sol's Ultrafast mode is currently available as a limited preview and will gradually be rolled out to more users. Among the earlier adopters of the GPT-5.6 Sol Ultrafast mode are quantitative trading firm Jane Street, AI lead generation company Podium, and financial companies Basis and Rogo.

Cerebras shares fell 11.85% in US trading on Thursday, having dropped nearly 15% intraday. Data shows that the company’s Q2 revenue rose 74% year-on-year to $180.1 million, missing analysts' average estimate of $194 million; the loss per share was $2.98, considerably worse than the analyst consensus of a $0.17 loss per share. By segment, second-quarter hardware revenue dropped 23% to $54.1 million year-on-year, indicating that this innovative chip architecture startup still faces challenges in commercialization. However, cloud and other service revenues hit a record high, growing 281% year-on-year to $126 million.

Since its initial public offering (IPO) in May, Cerebras' stock price has risen almost 29%. The core tech highlight of Cerebras lies in its unique high-end processor design—to manufacture a single chip from a silicon wafer usually used to make multiple components. Cerebras has built an extensive data center network and offers computing power leasing services, originally intended to verify the feasibility of its technology. Currently, with robust demand for AI computing power, these leasing services have become an important driver of the company’s growth.

Cerebras positions itself as a challenger to Nvidia in the AI chip sector, but its largest source of income now is cloud computing services. Cerebras CEO Andrew Feldman stated: "In terms of order delivery and revenue recognition, the hardware business will show volatility. That's inherent to the industry." He added that some customers are not yet ready with the data center space needed for new computing systems. He also emphasized that the company remains committed to selling hardware systems and data center services.

Looking ahead, Cerebras expects revenue of about $215 million in the third quarter, above the analyst average estimate of $212 million; core gross margin is expected to be 38% to 40%, also higher than the consensus of 36%. The company raised its full-year revenue guidance to $880–890 million, up from a prior forecast of $855–865 million, while analysts’ consensus is $867.6 million. The company also forecasts full-year core gross margin will be between 41% and 43%.