Perplexity has launched Sonar, an advanced AI answer engine built on Cerebras' hardware, allowing developers to achieve high-speed responses via an API. Unlike traditional search engines, Sonar targets integration into applications, enhancing performance in various products. This unique service boasts speeds reaching up to 1,200 tokens per second, close to Cerebras' record of 1,500 tokens. With four model variants, including reasoning options, Sonar caters to different querying needs. Overall, the initiative signifies an evolution in AI-driven search technology, emphasizing efficiency and adaptability.
The base Sonar model is a modified version of Meta's Llama 3.3-70B, designed to run on Cerebras' AI hardware for exceptional performance.
Sonar by Perplexity aims at developers, allowing them to generate answers rapidly via an API, distinguishing it from typical search engines.
Collection
[
|
...
]