Perplexity open-sources pplx-embed-v2-late multimodal embedding models
CShorten30 · x · 2026-10-08
Perplexity is releasing pplx-embed-v2-late, two late-interaction embedding models that retrieve text, images, and pages via a shared embedding space enabling cross-model querying. The team says both models achieve frontier retrieval performance and are publicly available on Hugging Face under an open license, positioning the family as a shift toward multimodal, multi-vector embeddings rather than just a performance bump.
Related event: Perplexity open-sources pplx-embed-v2-late retrieval models(36 posts)→
More from Multimodal
- Kokoro-82M distilled to an 8M-param TTS: just 9MB of weights, runs faster than real time on one CPU thread — multimodalart · 2026-10-08
- Google opens SynthID detector to the public as 180 billion images and videos carry watermarks — The Decoder · 2026-10-08
- Tiny vision models hit 17ms per webcam frame on a 4090, fast enough to play games — maximelabonne · 2026-10-08
- 17 AI comic tutorials analyzed: consistency is four problems, full-page gen fails — Ok_Low_5536 · 2026-10-08
- Minimax H3 ecosystem roundup: retro-fantasy LoRA, one-pass inpainting, and six new ComfyUI tools — optimisticalish · 2026-10-08
- ElevenLabs Brings 11 Speech Models to OpenRouter With 50% Off Until October 19 — lukeharries · 2026-10-08