Providers outside the index, such as the hyperscalers: each one's lowest on-demand price per GPU-hour as it posts it, not standardized to the benchmark contract; a price posted in euros is converted at the European Central Bank's rate of the day, with the euro price under it (methodology).
Qwen4 Max is a model from Alibaba. Its posted API price is $1.425 per 1M input tokens, $0.07125 per 1M cached input tokens and $4.275 per 1M output tokens, or $0.276 per 1M tokens at the mix of tokens agents send (87% cache read, 12% input, 1% output). It carries 21.8% of Alibaba's API tokens and is one of the 3 models in the Alibaba token price index, $0.129 per 1M tokens today. Of the 35 models in the token index it is the 13th most expensive. At today's H100 rental price of $3.78 per GPU-hour, its break-even throughput is 491 tokens per second per GPU at its list price (3 input : 1 output). Prices are collected once a day, for each UTC day.
Models near this price
Step 5 Preview (StepFun): $0.234 per 1M tokens at the mix, $1.00 input and $2.70 output
GPT-6.1 Terra (OpenAI): $0.213 per 1M tokens at the mix, $1.00 input and $5.00 output
Gemini 3.8 Flash (Google): $0.193 per 1M tokens at the mix, $0.75 input and $3.75 output
Mistral Large 4 (Mistral): $0.166 per 1M tokens at the mix, $0.70 input and $2.10 output
Qwen3.8 27B (Alibaba): $0.150 per 1M tokens at the mix, $0.425 input and $2.55 output
Gemini 3 Flash Preview (Google): $0.134 per 1M tokens at the mix, $0.50 input and $3.00 output
Cite as: Compute Screener, “Qwen4 Max API price”, computescreener.com/models/qwen4-max. The figures are free to reuse under CC BY 4.0 with credit to Compute Screener.