OpenRouter has added Qwen: Qwen4 Max to its model catalog. The listing offers access through OpenRouter’s API, which the service says is compatible with the OpenAI API used by many software tools. Developers can switch to the Qwen4 Max model name and OpenRouter’s base URL, though the page does not provide information about the model’s abilities or intended uses.
The posted rates are $1.50 per million input tokens and $4.50 per million output tokens. Tokens are small pieces of text that models process and generate. OpenRouter notes that caching and discounts can make the price customers pay lower than the posted rate, but its page does not give a reliable average price for this selection.
What the listing shows
OpenRouter says Qwen4 Max is hosted by one provider and that it forwards requests without routing them among providers. The page reports throughput of 46 tokens per second and a median round-trip latency of 0.89 seconds. These are service metrics shown by OpenRouter, not results from a model-quality test; the listing gives no benchmark scores or comparison with other models.
The page also shows 100% availability, while explaining that its uptime measure covers the past three days. Because the model is served by one provider, OpenRouter’s usual ability to send a failed request to another healthy provider may not apply to this listing. The page does not provide enough detail to establish how the reported availability will hold over a longer period.
There are no public apps using the model with enough traffic to appear in OpenRouter’s usage section. The listing also says Qwen4 Max is not available through EU in-region routing. That is a practical restriction for users who need requests processed within the EU; the page does not say when that routing option might become available.
Why it matters
Developers can call Qwen4 Max through OpenRouter using an API format compatible with many OpenAI-based tools, without switching to a different API style. But teams that require EU in-region processing cannot use this listing for that purpose, and the page does not provide model-quality tests or longer-term reliability data.