Perplexity AI has introduced two new online language models, `pplx-7b-online` and `pplx-70b-online`, which are designed to provide helpful, factual, and up-to-date responses. These models leverage knowledge from the internet to ensure the most recent information is incorporated into their answers. They can handle time-sensitive queries that offline models struggle with. Perplexity Labs provides access to these models through its LLM playground.
The performance of Perplexity’s online models was evaluated using criteria such as helpfulness, factuality, freshness, and holistic evaluation. The results showed that the online models outperformed OpenAI’s `gpt-3.5-turbo-1106` and Meta AI’s `llama2-70b-chat` on freshness, factuality, and holistic criteria. The models are accessible via the PPLX API, which is now in general release. Pricing for usage of the models is based on a new usage-based structure, and Pro users receive a monthly credit. Further information and resources are available on the Perplexity AI website.
HEALTH
WORLD AFFAIRS
TECHNOLOGY
CREATIVE
