创业邦·2025-11-07 00:09Group 1 - Keda Xunfei launched the "Transformable Voice Replication" technology based on the Spark Voice large model, allowing users to replicate any voice with high fidelity using just one recording, which is expected to revolutionize fields such as digital humans, audiobooks, and content creation [2] - Yushu Technology unveiled a full-body remote operation platform, demonstrating its G1 robot performing household tasks like washing dishes, mopping, organizing clothes, and serving water under human remote control [2] - Keda Xunfei released the Spark X1.5 deep reasoning large model, achieving over 93% efficiency compared to international competitors in end-to-end performance, with capabilities in language understanding, text generation, knowledge Q&A, logical reasoning, mathematical skills, and coding, while supporting over 130 languages and reaching 95% performance of GPT-5 [2] - Google's AI model series Gemini made significant progress with the preview version "gemini-3-pro-preview-11-2025" discovered on the Vertex AI platform, indicating a release this month with support for a massive context window of up to 1 million tokens [2][3] Group 2 - The preview model identifier "11-2025" clearly points to a release window in November, with Vertex AI being a core tool for Google's cloud AI development platform, capable of handling up to 200,000 tokens at standard level and expanding to 1 million tokens at advanced level [3]