CryptoWorld News: Doubao doubao-seed-2.0-lite has been upgraded to a fully modal understanding model. The lite version outperforms the pro version on multiple benchmarks, with speech recognition performance surpassing Gemini 3.1 Pro. The model simultaneously handles video, images, audio, and text, supporting speech transcription in 19 languages and machine translation between 16 language pairs. In terms of vision, the new lite version exceeds doubao-seed-2.0-pro on advanced academic benchmarks such as physical reasoning and medical question answering, and achieves state-of-the-art results in fine-grained perception and embodied understanding. The model is compatible with frameworks like openclaw and hermes agent, enhancing multi-step task decomposition and long-range task stability, supporting cross-application continuous business process execution.