Perplexity releases pplx-embed-v2-late: a 0.6B edge model and a 9B model scoring 92.4% on MADQA
Perplexity has released two new multimodal embedding models that retrieve text, images and PDF pages in a single vector space. The 0.6B version targets laptops and edge devices, while the 9B version is built for maximum quality and reached 92.4% accuracy on the MADQA benchmark. Both are available on Hugging Face under the MIT license, though a hosted API is not live yet.
