Apple Trains Dedicated LLM for China with Alibaba's Help, New Regional AI Architecture Revealed
APPSO · wechat · 2026-08-15
Apple has developed a dedicated large language model for the Chinese market in partnership with Alibaba, which provided training support, according to a Reuters exclusive report. This suggests that the Chinese version of Apple Intelligence may integrate capabilities from Apple's proprietary model, Alibaba's Qwen, and Baidu's technology.
Technically, the model likely adopts the AFM3 family architecture. AFM3Core serves as an on-device dense model, while AFM3CoreAdvanced uses a sparse activation architecture with Instruction-Following Pruning to dynamically load expert weights between NAND flash and DRAM, optimizing performance and memory usage. Cloud models cover general tasks, image generation, and complex reasoning, though the deployment location and privacy framework for China remain unconfirmed.
Apple has passed China's regulatory filing and expects to launch the features via an iOS update in the coming months. This regionalized architecture marks a shift from a unified global experience, potentially leading to significant differences in model capabilities, cloud services, and data processing rules across regions.
Related event: Apple Partners with Alibaba for China-Specific AI Model(7 posts)→
More from Companies & People
- ai& × AMD Build Day hackathon in Tokyo: 8 teams, 4.5 hours, Chinese models allowed — DavidBennett__ · 2026-10-16
- Sakana AI opens first dedicated ISV, SI and CS roles in Japan GTM push — SakanaAILabs · 2026-10-16
- Kaggle teams with Google DeepMind and Zapier for inaugural Bench Summit — jocarrasqueira · 2026-10-16
- Ethan Mollick jokes people take his shipping-timeline tweet too seriously — emollick · 2026-10-16
- Mantic AI cofounders on building AI that forecasts world events better than top humans — tshevl · 2026-10-16
- Anthropic readies work-email discount path to annual Claude Pro, checkout spotted — ryanmerket · 2026-10-16