Alibaba is going all in on Qwen…

Alibaba announces $52B three-phase roadmap to ASI at Apsara 2025, releasing Qwen 3 Max with 1T+ parameters.

• Alibaba committed $52B across three phases targeting ASI by 2032, framing themselves as the Android of the AI era.
• Qwen 3 Max base model exceeds 1 trillion parameters, pre-trained on 36 trillion tokens using mixture-of-experts architecture.
• Qwen 3 VL tops ClockBench (new vision benchmark) at 39% analog clock accuracy, released open-source and open-weight.
• Qwen 3 Omni is a multimodal model handling vision, audio, text, and speech simultaneously.
• Alibaba’s three phases: generalized understanding (past), autonomous tool use (now), self-iterating physical-world AI (2032 target).
• A reasoning variant of Qwen 3 Max Instruct is currently in training and benchmarks competitively against thinking models.

2025-10-03 · Watch on YouTube