Alibaba has announced a full-stack AI roadmap at its Apsara Conference covering chips, cloud infrastructure, models and agents. The company said its next-generation Qwen 4 model is in training and outlined plans for Qwen 4.5 and Qwen 5 series that could scale to 5–10 trillion parameters. Multimodal updates include Qwen3.8-LiveTranslate for simultaneous interpretation, a new audio-generation model, upgraded speech recognition and text-to-speech systems, and Qwen-Image 3.1 for design and e-commerce. T-Head unveiled the Zhenwu V900 AI processor with 216 GB memory and three times the performance of its predecessor, scheduled for commercial release in Q1 2027, plus a Yitian CPU roadmap for agentic workloads. Alibaba Cloud introduced an agentic cloud built around model, harness and context layers, including AgentCore for lifecycle management of agents, an Agent Security Center, and Agent Context services that can cut token usage by up to 67%.
Alibaba has announced a full-stack AI roadmap at its Apsara Conference covering chips, cloud infrastructure, models and agents. The company said its next-generation Qwen 4 model is in training and outlined plans for Qwen 4.5 and Qwen 5 series that could scale to 5–10 trillion parameters. Multimodal updates include Qwen3.8-LiveTranslate for simultaneous interpretation, a new audio-generation model, upgraded speech recognition and text-to-speech systems, and Qwen-Image 3.1 for design and e-commerce. T-Head unveiled the Zhenwu V900 AI processor with 216 GB memory and three times the performance of its predecessor, scheduled for commercial release in Q1 2027, plus a Yitian CPU roadmap for agentic workloads. Alibaba Cloud introduced an agentic cloud built around model, harness and context layers, including AgentCore for lifecycle management of agents, an Agent Security Center, and Agent Context services that can cut token usage by up to 67%.