✳ the wire · analysis

Qwen 4 Max is out: $1.50 / $4.50 per million tokens and a 1M-token context, on Alibaba's API

Qwen 4 Maxconfirmedby ArtificialWatch

An email when every new Alibaba model answers on a public API, about 15 minutes after members, plus every frontier launch · no card

ArtificialWatch title card · verified against a primary source

Alibaba released Qwen 4 Max on November 16. Its Qwen account announced Qwen4-Max at 02:01 UTC with the API live on QwenCloud as qwen4-max, and OpenRouter listed it at the same price by 02:18 UTC. It is the first Qwen 4 model with a pay-as-you-go price, five days after a two-week Qwen4-Max-Preview reached Qwen Chat.

THE PRICE

  • $1.50 per million input tokens and $4.50 per million output tokens (QwenCloud's model page, read Nov 16).
  • $0.075 per million for implicit-cache input. Explicit cache creation costs $1.875 per million for five minutes or $3 for an hour, and explicit cache reads cost $0.075.
  • Alibaba says all three headline prices are below Qwen3.8-Max's, with implicit caching down 70%.

THE SPECS

  • A 1M-token context: up to 991K input tokens and 393K output tokens, with up to 262K reasoning tokens.
  • Text, image and video in, text out. OpenAI- and Anthropic-compatible APIs, function calling, structured outputs, web search, batch jobs and fine-tuning (QwenCloud).
  • Alibaba calls it a new mixture-of-experts architecture. The Max weights are not public; Alibaba says open weights of smaller Qwen4 models will follow.

ALIBABA'S OWN NUMBERS

  • Alibaba says Qwen4-Max beats Qwen3.8-Max while using up to about 50% fewer tokens, and that average cost per task fell by more than 60% in its internal evaluation.
  • Alibaba's chart compares it with Qwen3.8-Max, Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra and Kimi K3.1 on 16 benchmarks. Against Opus 5.5 it is ahead on six, including DeepSWE v1.1 (75.7 vs 74.2), Agents' Last Exam v1.1 (46.7 vs 43.4) and HLAB-Vals (21.3 vs 9.2), and behind on ten, including Terminal-Bench 4.0 (48.5 vs 59.6), ProgramBench (31.5 vs 65.0) and Humanity's Last Exam (51.7 vs 59.8).
  • These are Alibaba's figures. No independent leaderboard had scored it when this was written.

HOW WE CAUGHT IT

ArtificialWatch's catalog sweep first saw qwen4-max at 02:01:05 UTC and fired its alert at 02:02:07 UTC. Alibaba's post went up at 02:01:55 UTC.

WHAT WE DO NOT KNOW

How it scores on independent benchmarks, when the smaller open-weight Qwen 4 models arrive, and when Qwen 4 Plus and Qwen 4 Flash follow. Alibaba has given no dates for any of them.

Source: @Alibaba_Qwen on X · QwenCloud model page, read directly · OpenRouter ↗ · Qwen 4 Max tracker · the bench index
Data: Open-weights leaderboard · LLM API pricing

Cite this page

Writing about this story? Copy a link or a citation.

<a href="https://artificialwatch.com/wire/qwen-4-max-is-out-1-50-4-50-per-million-tokens-1m-context">Qwen 4 Max is out: $1.50 / $4.50 per million tokens and a 1M-token context, on Alibaba's API</a>
ArtificialWatch, "Qwen 4 Max is out: $1.50 / $4.50 per million tokens and a 1M-token context, on Alibaba's API", https://artificialwatch.com/wire/qwen-4-max-is-out-1-50-4-50-per-million-tokens-1m-context
sweeping every 60 seconds

Hear when every new Alibaba model drops — not when we write it up.

Claude Opus 5 went live at 16:51 UTC. Members had the alert at 16:52. Free gets a Chrome push and an email about 15 minutes after members.

  • Free forever
  • No card
  • Unsubscribe in one click

Want it the minute it answers? Try Signal 7 days free →

← back to the wire · every story, by month