Feb 11, 07:39 PM

Perplexity AI Launches Sonar Model on Llama 3.3 with 70 Billion Parameters, Achieving 1,200 Tokens/Second

Perplexity AI has introduced its new AI model, Sonar, which is built on the Llama 3.3 architecture with 70 billion parameters. The model reportedly outperforms existing models such as GPT-4o-mini and Claude 3.5 Haiku in user satisfaction while matching or exceeding the performance of top models like GPT-4o and Claude 3.5 Sonnet. Sonar is optimized for both speed and answer quality, achieving an impressive processing rate of 1,200 tokens per second, which is ten times faster than comparable models. This advancement is powered by Cerebras Systems' inference technology, marking a significant enhancement in AI-driven search technology. Perplexity AI's retention metrics indicate positive user reception of the new model, which is expected to transform how information is retrieved and consumed online.

#AI #Sonar #Llama #Cerebras Systems

Written with ChatGPT (GPT-4o mini).

Perplexity AI Launches Sonar Model on Llama 3.3 with 70 Billion Parameters, Achieving 1,200 Tokens/Second

Sources

Additional media

Similar Stories