Speeding up to 14 times! Cerebras (CBRS.US) has created a super-fast mode for OpenAI's GPT-5.6 Sol model.

Artificial intelligence (AI) chip newcomer Cerebras Systems (CBRS.US) announced on Thursday that it is offering a brand-new ultrafast mode for OpenAIs GPT-5.6 Sol model. Reports indicate that this new ultrafast mode can process data up to 14 times faster than OpenAI's standard mode and is currently being rolled out gradually to a select group of customers who require lower latency. Cerebras stated on its website: The ultrafast mode is initially being made available to a small number of customers, with plans to gradually expand access in the future. Cerebras provides computational support for the ultrafast mode of GPT-5.6 Sol, achieving a speed of up to 750 output tokens per second without sacrificing quality, thus enabling Sol ultrafast to accelerate processing for the most time-sensitive and mission-critical tasks. OpenAI noted in another article: Their early usage is helping us understand in which scenarios a tenfold speed increase can create the most value, and what changes occur when the model can keep up with the user's speed. As computational capacity continues to grow, we will leverage these findings to guide deployment. OpenAI also mentioned that it has identified some "encouraging" application scenarios for the ultrafast mode, including event response and reliability, financial research and security, customer support and voice, business, as well as real-time research and experimentation. The ultrafast mode of GPT-5.6 Sol is currently available in a limited preview and will be gradually expanded. Some of the early users of the GPT-5.6 Sol ultrafast mode include quantitative trading firm Jane Street, AI lead generation company Podium, as well as financial firms Basis and Rogo. On Thursday, Cerebras' shares fell 11.85% in U.S. trading, with an intraday decline of nearly 15%. Data shows that the companys revenue for Q2 grew 74% year-over-year to $180.1 million, falling short of analyst expectations of $194 million; the company reported a loss of $2.98 per share, significantly worse than the analyst consensus of a loss of $0.17 per share. By business segment, second-quarter hardware revenue declined 23% year-over-year to $54.1 million, indicating that this innovative chip design startup still faces challenges in its commercialization process. However, revenue from cloud and other services reached a record high, increasing by 281% year-over-year to $126 million. Since its initial public offering (IPO) in May, Cerebras stock price has risen nearly 29%. The core technological selling point of Cerebras lies in its unique high-end processor designa single chip made from silicon wafers that are typically used for manufacturing multiple components. Cerebras has established a vast network of data centers and offers computational power leasing services, initially intended to validate the feasibility of its technology. Currently, the strong demand for AI computational power has instead made the leasing service a significant growth engine for the company. Cerebras positions itself as a challenger to NVIDIA Corporation in the AI chip sector, but its largest revenue source now comes from cloud computing services. Cerebras CEO Andrew Feldman stated: From the perspective of order delivery and revenue recognition timelines, the hardware business will exhibit volatility. This is a characteristic of the industry. He added that some clients are currently not prepared to accommodate the data center space required for the new computing systems. He also emphasized that the company remains committed to selling hardware systems and data center services. Looking ahead, Cerebras expects Q3 revenue to be around $215 million, higher than the analyst average estimate of $212 million; core gross margins are expected to be between 38% and 40%, also exceeding the analyst average estimate of 36%. The company has raised its full-year revenue guidance to $880 million to $890 million, up from the previous outlook of $855 million to $865 million, while the analyst consensus stands at $867.6 million. The company also expects its core gross margin for the year to fall between 41% and 43%.