OpenAI Unveils New AI Model in Collaboration with Cerebras Systems
OpenAI is set to launch its inaugural artificial intelligence (AI) model utilizing semiconductor technology from Cerebras Systems Inc., signaling a strategic endeavor to diversify its partnerships beyond the established domain of Nvidia Corp.
The newly developed model, designated GPT-5.3-Codex-Spark, represents a swifter, albeit less potent, adaptation of the recent Codex software aimed at streamlining coding automation.
The Spark model, anticipated for release on Thursday, empowers software engineers to efficiently undertake tasks such as modifying specific segments of code and executing tests.
Notably, it permits users to interrupt the process seamlessly, allowing for additional coding requests without enduring the delays associated with extensive computational activities.
In a significant maneuver last month, OpenAI secured a contract exceeding $10 billion for hardware from Cerebras, enhancing the responsiveness of its AI models. This collaboration marks a pivotal moment for Cerebras, providing it with a substantial advantage in a landscape traditionally ruled by Nvidia.
For OpenAI, this agreement signifies its relentless pursuit of a more varied supply chain to cater to burgeoning computational demands. Notably, in October of the previous year, OpenAI finalized a landmark deal with AMD, acquiring a staggering 6 gigawatts of AMD graphics processing units over several years.
In the same month, OpenAI also committed to purchasing customized chips and networking components from Broadcom Inc.
Recently, however, the dynamic between OpenAI and Nvidia has faced scrutiny due to reports of underlying tensions. Nevertheless, the CEOs of these firms have publicly reaffirmed their commitment to ongoing collaboration.
In a statement issued the previous Thursday, an OpenAI representative reiterated the foundational nature of the partnership with Nvidia, emphasizing that the organization’s most advanced AI models are a result of extensive “multi-year hardware and software engineering” done in unison.
“This is why we continue to anchor our training and inference stack on Nvidia while deliberately broadening our ecosystem through alliances with Cerebras, AMD, and Broadcom,” the spokesperson articulated.
This new model epitomizes OpenAI’s latest initiative to eclipse competitors such as Alphabet Inc’s Google and Anthropic PBC, who are vying for supremacy in the burgeoning market for AI-enhanced coding assistants. OpenAI reported that Codex boasts over a million active users weekly.
Initially, the enhanced AI model will be accessible to ChatGPT Pro subscribers as a research preview, with intentions to extend availability to a broader audience in the ensuing weeks.
Source link: Taipeitimes.com.
