Chinese labs shipped 7 major models in 3 weeks. All under $1/M input tokens. Can Western labs justify 5-10x pricing?
Seven major models from Moonshot, ByteDance, and Alibaba all priced below $1 per million input tokens.
Deep Dive
Chinese AI labs including Moonshot AI, ByteDance, and Alibaba released seven major models in three weeks, all priced under $1 per million input tokens. Key releases include Kimi K2.5 (coordinates 100 sub-agents), GLM-5 (lowest hallucination rate), and Qwen 3.5 (claims to beat GPT-5.2 on 80% of benchmarks). Four of five text models are open-weight, all use MoE architectures, and all are built for agent-based workflows, not just chatbots.
Why It Matters
This pricing pressure forces Western labs to justify 5-10x higher costs and could accelerate global adoption of cheaper, open-weight AI agents.