Join our daily and weekly newsletters for the latest updates and the exclusive content on AI coverage. Learn more
Brain systems And Perplexity Ai join forces to challenge the domination of conventional search engines, announcing a partnership which promises to provide research results fueled by the IA almost instantaneous at previously deemed speeds impossible.
The collaboration, announced in an exclusive report by Venturebeat, focuses on the new perplexity Sonar modeldeer Specialized AI chips At 1,200 tokens per second – which makes it one of the fastest IA research systems available. Built on meta- LAMA 3.3 70B Foundation, Sonar represents an important bet that users will adopt research experiences on AI-STIs if they are quite fast.
“Our partnership with Cerebras has helped bring Sonar to life,” said Denis Yarats, CTO de Perplexity, in a press release. “The Infrastructures of inference of the Cerebras cutting edge AI allowed us to reach unprecedented speeds and efficiency.”
AI research has become faster – and Big Tech should be careful
The timing is notable, coming only a few days after the forms made the headlines with its Deep implementationwhich has demonstrated speeds 57 times faster than traditional solutions based on GPU. The company seems to be taking advantage of this momentum to establish itself as the essential supplier for high -speed inference.
According to the internal tests of Perplexity, Sonar surpasses both GPT-4O Mini and Claude 3.5 Haïku “By substantial margin” in user satisfaction measurements, while corresponding or exceeding more expensive models like Claude 3.5 Sonnet. The company’s assessments show that SONAR reaches billing scores of 85.1 out of 100, against 83.9 for GPT-4O and 75.8 for Claude 3.5 SONNET.
Specialized material: the new battlefield for IA companies
The partnership reflects an increasing trend in AI companies in search of competitive advantages thanks to specialized equipment. CEO of Cerebras, Andrew Feldman, recently argued that such technological advances are developing rather than contracting the market. “Whenever the calculation has been made cheaper, they [public market investors] have systematically supposed that the market is smaller. Feldman said to Zdnet in a recent interview. “And in each case, more than 50 years, he has enlarged the market.”
Industry analysts suggest that this alliance could put pressure on traditional and other research providers AI companies To reconsider their material strategies. The ability to provide almost insufficient results could be particularly convincing for corporate customers, where speed and precision have a direct impact on productivity.
Can market impact: Can specialized fleas reshang business research?
However, questions remain on the scalability and profitability of specialized AI chips compared to traditional solutions based on GPU. Although Cerebras has shown impressive speed advantages, the company is confronted with the challenge of convincing customers that social benefits justify potential premium prices.
The partnership also highlights the increasingly competitive landscape in AI research, where companies run to differentiate by speed and precision rather than the size of the gross model. For perplexity, which drew attention as a native alternative to traditional search engines, Cerebras partnership could help establish it as a serious competitor on the business research market.
Perplexity plans to make Sonar available to professional users initially, with greater availability to come soon. Companies have not disclosed the financial terms of their partnership.