BREAKING
US Startups Adopt Cheaper Chinese AI
Inference costs spark 'tokenmaxxing'
Two competing strategies
US Players
Frontier
●
OpenAI and Anthropic
●
Top-end performance
●
Higher inference costs
Chinese Players
Open-weight
●
DeepSeek, Qwen, GLM
●
Far lower prices
●
Self-hostable
Three models drawing attention
1
DeepSeek V4, MIT
↓
2
Alibaba Qwen, Apache 2.0
↓
3
Zhipu GLM-5.2, MIT
0
Terminal-Bench 2.1
0
M
token context
Comparable performance, lower cost
AI NEWS BLITZ
A growing number of US startups are switching to cheaper Chinese AI models.