DeepSeek new model/GLM-5.2/MiniMax M3/Qwen3.7/TCO
Key Questions
Which Chinese AI models are outperforming or competing with Western ones on cost?
GLM-5.2 tops CodeArena at 1/6 the cost of GPT-5.5, with an 80% switch by US companies due to a 43x cost gap and only 2.7% performance delta. Other models include MiniMax M3 and Qwen3.7.
What concerns exist around Kimi K2.7 Code in recent reviews?
Kimi K2.7 Code received a scathing review highlighting regressions, despite the broader trend of Chinese models gaining enterprise traction on cost and performance.
How are Chinese AI models impacting US developers and startups?
Lower API prices and open weights make Chinese models harder to ignore, with 80% of US companies switching amid the significant cost advantages and minimal performance trade-offs.
GLM-5.2 tops CodeArena, 1/6 cost vs GPT-5.5, dual listing. 80% US companies switched to Chinese models (43x cost gap, 2.7% performance delta). Kimi K2.7 Code scathing review—regressions. Xiaomi releases AI coding tool. Zhipu slime v0.3.0 for agent RL. Anthropic Fable 5 exit opens door.