Jina-CLIP v2, a 0.9B multimodal embedding model with multilingual support of 89 languages, high image resolution at 512x512, and Matryoshka representations.