Qwen 4 is Alibaba's next-generation AI model family, and only one model is usable today: Qwen4-Max-Preview. It has a 1M-token context window, runs only on Alibaba's Token Plan subscription from $6 a month, and ends on November 25, 2026. The full Qwen 4 family has no release date. This guide covers specs, pricing and what is still unconfirmed, so you can decide whether to try it now or wait.

Key takeaways

• What it is: The first model of Alibaba's Qwen 4 generation, a multimodal reasoning model with text, image and video input.
• Release date: The preview opened on November 11, 2026 and is retired on November 25. The full Qwen 4 release date is not announced.
• Price: No per-token price. Access is through Token Plan credits, from $6 a month ($3 for new users until November 30).
• Benchmarks: None published. Alibaba has also not released weights, a model card, a parameter count or a license.
• On 1min.AI: Qwen 4 is not integrated yet. Build your test baseline with Claude Opus 5.5 and Gemini 4 Argon now.

What Is Qwen 4?

Qwen 4 is the next major generation of Alibaba's Qwen models. Alibaba launched Qwen in April 2023 as Tongyi Qianwen, and the line now covers text, vision, audio and math. Its most recent stable flagship is Qwen3.8-Max, released on August 3, 2026. 

Alibaba confirmed on September 22, 2026 that Qwen 4 was in training and projected that later Qwen 4.5 and Qwen 5 models would reach 5–10 trillion parameters.

Qwen 4 Release Date and Public Availability

The public release date of the full Qwen 4 family is unannounced. Only the Max preview is live, and the table below separates confirmed facts from open questions.

Alibaba says the preview model will not change during the window and will be retired when it ends. Any tool you connect to qwen4-max-preview will stop working after November 25.

What Qwen4-Max-Preview Offers

• Context: 1M tokens.
• Output: up to 393,216 tokens, shared between reasoning and the final answer.
• Input: text, images and video; the model returns text.
• Reasoning: thinking is on by default, with adjustable effort levels.
• API formats: compatible with the OpenAI and Anthropic API formats, according to its documentation.
• Model ID:qwen4-max-preview.

Qwen 4 Pricing and Access

Qwen 4 has no per-token price. Alibaba sells the preview only through its Token Plan, a monthly credit subscription, and pay-as-you-go is not supported. During the preview, Qwen4-Max uses credits at as little as 20% of the standard rate, so your quota lasts longer. Alibaba has not published the standard credit rate for this model.

You can also try the model in Alibaba's own apps, including the Qwen app and QwenStudio. One reviewer reports that Token Plan terms restrict use in app backends, so read the terms before connecting it to a product.

Qwen4-Max-Preview vs Qwen3.8-Max

Qwen4-Max-Preview keeps the 1M-token context of Qwen3.8-Max and raises the maximum output. We found no verified benchmark comparing the two, so the table covers specs only.

Because quality claims about Qwen 4 come only from the Qwen team, test it on your own prompts before you rely on it.

Qwen 3.8 Max vs Qwen 3.8 Flash: Which Should You Use?

Qwen 3.8 Max is Alibaba's most capable hosted model, while Qwen 3.8 Flash is the cheaper, faster option for high-volume work. Both are current stable models, so they are the practical choice today while Qwen 4 is still in preview. The main trade-off is capability against cost.

Output on Flash costs roughly 13 times less than on Max. That gap grows in agent workflows, where the same long context is sent again and again and cached input is the main cost.

When Qwen 3.8 Flash Is the Smarter Choice

• You run high-volume tasks, such as classification, summaries or support replies.
• Your agents resend long context, so cached-input price matters.
• You need speed and a low bill more than top reasoning quality.

When Qwen 3.8 Max Is Worth the Extra Cost

• You work on the hardest reasoning and coding tasks.
• Answer quality matters more than cost per request.
• You need the flagship's thinking and non-thinking modes.

Prices vary by provider and region, and a few trackers list different Max rates. Confirm current prices on Alibaba Cloud before you budget. Scores are from one independent index, so test both models on your own prompts.

How to Prepare for Qwen 4

A short test plan makes the preview useful, and it still pays off if you wait for the full release.

1. Define your use cases. List the tasks you would test first, such as long-document analysis, coding or content drafting.
2. Build a small test set. Write 10 to 20 real prompts and record how your current model answers them.
3. Try the preview before November 25. A $3 to $6 Token Plan is enough for a quick test of the 1M-token context.
4. Track official channels. Watch Alibaba's Qwen account, Alibaba Cloud documentation and the Qwen organization on Hugging Face for the full release.

Test Frontier Models on 1min.AI While You Wait

A baseline is easier to build in one workspace. On 1min.AI, you can run the same prompt on Claude Opus 5.5 and Gemini 4 Argon with Multi AI Chat and compare the answers side by side. Qwen 4 is not available on 1min.AI yet, but the results you collect now give you a baseline to score it against later.

Developers can also reach several model families through one API. Read the Gemini 4 Argon API guide, or see how Gemini 4 Carbon compares with Claude Opus 5.5.

Chat With Claude Opus 5.5 and Gemini 4 Argon

Qwen 4 FAQs

Is Qwen 4 available to the public?

Partly. Qwen4-Max-Preview is open to Token Plan subscribers from November 11 to November 25, 2026. The rest of the Qwen 4 family is not available.

When is the Qwen 4 release date?

Alibaba has not announced one. Its September 22, 2026 statement says only that Qwen 4 is in training. Dates from conference attendees are unverified.

How much does Qwen 4 cost?

There is no per-token price. The preview runs on Token Plan credits, from $6 a month or $3 for new users until November 30.

What is the context window of Qwen 4?

Qwen4-Max-Preview supports 1M tokens of context and up to 393,216 output tokens. Other Qwen 4 models have no published specs.

Is Qwen 4 open source?

Unknown. Alibaba has published no weights or license for any Qwen 4 model.