Kimi K2.6 surpassed Opus 4.5 with a score of 56.3 in 4.8 months, and GLM-5.2 cleared GPT-5.2 with a score of 72.4 in 6 months.
这两个数字是判断模型层会否商品化的关键。但分子分母都由作者自选：换一组基准、换一个时代起点模型，追赶期就会变。当成方向性信号可以，当成定量结论会踩坑。
4 Matching Annotations
- Last 7 days
-
newsletter.semianalysis.com newsletter.semianalysis.com
-
-
www.apple.com www.apple.com
-
This creates a vast shared memory pool across systems, allowing users to load the largest and most demanding frontier-class open-weight models available today.
用 Thunderbolt 5 加 RDMA 把多机拼成共享内存池，是在没有 NVLink 的前提下硬凑容量。注意苹果只敢报四台三倍，已见明显衰减。这条路线适合权重长期驻留的本地推理，不适合训练，也撑不起多租户服务。
-
- Feb 2026
-
-
Airbnb picks Alibaba’s Qwen over ChatGPT in a win for Chinese open-source AIAirbnb ‘relies heavily’ on Alibaba’s Qwen models to power its AI customer service agent, CEO Brian Chesky says
Even US firms are using Chinese models because they are so much cheaper. They do not use them for everything, but where they can get away with lower cost options they will.
-
-
www.chathamhouse.org www.chathamhouse.org
-
Low-cost Chinese AI models forge ahead, even in the US, raising the risks of a US AI bubble Nvidia’s latest earnings report reassured some. But Chinese AI models are fast gaining a following around the world, underlining concerns over an ‘AI bubble’ centered on high-investment, high-cost US models.
-