Chinese AI Models Match GPT-4o at Fraction of Cost

A developer tested four Chinese AI models—DeepSeek, Qwen, Kimi, and GLM—via a unified API and found them comparable to GPT-4o for many tasks. DeepSeek V4 Flash costs $0.25 per million output tokens, 40x cheaper than GPT-4o. The models offer 128K context windows and work with OpenAI's API format.