#multimodal-llm

#multimodal-llm

Moonshot's open-source Kimi K2.5 is a powerful multimodal coding model that can generate front-end web interfaces from images or video and matches leading coding benchmarks.

Google released Gemini 3 Flash: a faster, cheaper multimodal model that matches top models on benchmarks and becomes the default in Gemini app and search.

OmniVinci is a multimodal LLM that integrates text, vision, audio, and robotics data to improve cross-modal perception and reasoning.