The Billion Dollar AI Race Just Broke
Qwen 3.8 Max challenges OpenAI and Anthropic at 5-10x lower API pricing
“it sat there thinking for 16 days, starting from an empty folder, writing, testing, and repairing its own code”
Qwen 3.8 Max is positioned as a frontier-quality multimodal model with 1M token context, agentic capabilities, and API pricing 5-10x cheaper than OpenAI and Anthropic, with weights to be released publicly. The model demonstrated 16 days of autonomous coding from scratch, and open models now score over 50% on Humanity's Last Exam — up from 2% a year ago for top closed models. The pricing pressure from Qwen and DeepSeek may force incumbent labs to cut prices.