Alibaba's Qwen4-Max Goes Live at $1.50 a Million Tokens
Alibaba moved Qwen4-Max from preview to general release with a public API priced at $1.50 per million input tokens and $4.50 per million output tokens. That's 25 percent below its own Qwen3.8-Max on both sides, with cached input down 70 percent to $0.075, a 1 million-token context window, and a maximum output of 393,000 tokens.
The model is the headline, but the price is the strategy. Alibaba cut its flagship's token prices by a quarter, cut cached input by 70 percent, and put the result on every channel it owns within hours.
Alibaba launched Qwen4-Max properly overnight, a week after its preview. The general release adds the thing the preview didn't have: a public pay-as-you-go API, priced at $1.50 per million input tokens, $4.50 per million output tokens, and $0.075 per million cached input tokens, according to the model's QwenCloud pages.
The comparison that matters is against Alibaba's own last flagship. Qwen3.8-Max charges $2.00 input and $6.00 output per million tokens, with cached input at $0.25. Qwen4-Max is 25 percent cheaper on input and output and 70 percent cheaper on cached input, and its maximum output rises to 393,000 tokens from 131,000. The context window stays at 1 million tokens.
The model itself is a mixture-of-experts system with native vision, so image input isn't a separate product. Alibaba puts the parameter count at 2.4 trillion. Its own scorecard compares Qwen4-Max against Qwen3.8-Max and rival flagships, and those are vendor numbers: Alibaba ran them, on its own chosen tests, and they should be read that way until independent testing catches up.
Distribution moved as fast as the release. Within hours, the model was live on the QwenCloud API, in Qwen Studio, QwenWork and Qoder, and on OpenRouter and OpenCode Go. The endpoints are OpenAI- and Anthropic-compatible, which is the practical detail: a team already wired to either API shape can test Qwen4-Max by changing a model ID, not a codebase.
The preview route still exists, on a clock. Qwen4-Max entered preview on November 11 through Alibaba's Token Plan subscription only, at Individual tiers running $6 to $43 a month for 20,000 to 200,000 credits. The preview model ID bills at 20 percent of its standard credit rate, and 8 percent overnight Beijing time, until the preview is retired on November 25.
That preview exposed the one number Alibaba still hasn't published: the standard credit rate itself. Without it, a Token Plan subscriber can't compute what 20,000 credits actually buys in Qwen4-Max work. The pay-as-you-go API doesn't have that problem. Its price is a price, in dollars, per million tokens, which is why the general release is the real launch.
There's no downloadable version. Qwen4-Max is access, not weights: the API, the subscription plans, and Alibaba's own apps are the whole product. Teams that need a model running on their own hardware aren't this release's audience, and Alibaba isn't pretending otherwise.
Independent verification lags the launch. The Daily Fledgling's morning roundup noted that independent boards put the preview at the top of China's models, but the benchmark table shipping with the general release is Alibaba's own, run against a field that includes its in-house predecessor and rival flagships it selected. The specs and prices are checkable facts. The performance ranking is a claim, and it stays one until testers outside Alibaba reproduce it.
What the release does, immediately, is reset a price floor. A flagship-class model with a 1 million-token context now lists at $1.50 and $4.50. Every competing sales conversation this week starts from those two numbers, whether the competitor mentions them or not.
Sources
- How to Use Qwen4-Max: Every Way to Get Access Right Now — Apidog, November 16, 2026
- Qwen4-Max vs Qwen 3.8: Price, Context, Benchmarks and What Changed — Apidog, November 16, 2026
- Asia Tech News Digest — November 16, 2026
- The Daily Fledgling — Morning Edition, November 16, 2026
- Qwen4-Max is in preview until November 25. Where to try it, and what it costs — enjoy.dev, November 2026
The conversation.
Loading comments…

A little more perspective.
Reading is free. Verified members can comment and reply for $2 a month.