Short answer
Qwen 4 is not out yet, and there is no release date. As of November 7, 2026, Alibaba has not released a model called Qwen4, and there is no Qwen4 model on the official Qwen account on Hugging Face. Two things are confirmed: the official Qwen account posted a teaser on November 6, and Alibaba mentions Qwen4 in the Qwen3.8-Flash-Next model card.
- Official teaser: on November 6 at 12:00 UTC (20:00 Beijing time), the official Qwen account on X, @Alibaba_Qwen, posted a poster reading 北冥有鱼 其名为_ ("In the northern sea there is a fish, its name is _") with a single line of text: "The becoming begins." (post). The post doesn't mention Qwen4 and gives no date.
- Where Alibaba mentions Qwen4: the model card of Qwen3.8-Flash-Next, open-sourced on August 24, calls it "an experimental preview of the architecture that will underpin Qwen4" (Hugging Face).
- Leak, not officially confirmed: TestingCatalog posted on X that it expects Qwen4-Max, Qwen4-Flash, Qwen4-Plus and Qwen4-27B (post).
- What you can use now: Qwen3.8-Max (
qwen3.8-max) and Qwen3.8-Flash (qwen3.8-flash) are live on TeamoRouter behind an OpenAI-compatible API, with one key (Qwen3.8-Max model page, Qwen3.8-Flash model page).
New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.
Before Qwen 4 ships, call Qwen3.8 with the same key
What Alibaba has said
The November 6 teaser
The official Qwen account posted on X on November 6, 2026 at 20:00 Beijing time. The text is just "The becoming begins." The image is a portrait poster: the Qwen logo, a large flock of birds in the sky, and two lines of text, 北冥有鱼 其名为_ and "The becoming begins."
北冥有鱼 comes from the Zhuangzi, a classical Chinese text: "In the northern sea there is a fish, its name is Kun." The poster replaces the name with a blank. Alibaba hasn't explained the poster, named the new model or given a date, and the post doesn't say Qwen4.
The Qwen3.8-Flash-Next model card
Alibaba open-sourced Qwen3.8-Flash-Next on Hugging Face on August 24, 2026. The official model card calls it an experimental preview of the architecture behind Qwen4 and lists what changes in that architecture:
- Hybrid attention: Gated DeltaNet paired with a new Qwen Sparse Attention (QSA), which works on small blocks instead of single tokens to cut long-context latency.
- Gated Residual: gates on widened residual streams.
- N-gram Embedding: embeddings indexed by short n-grams, as another way to scale parameters.
- A new training recipe that applies the Muon and AdamW optimizers to different weight categories.
Qwen3.8-Flash-Next itself has 125B parameters with 6B active (plus 51B of n-gram embeddings and a 4B MTP layer), and a native context of 262,144 tokens, extensible to 1M. The same card says Qwen3.8-Flash is the official version based on Qwen3.8-Flash-Next, with a 1M-token context by default and official built-in tools.
So by Alibaba's own account, Qwen4 will be built on this architecture. Qwen4's own size, context and price haven't been announced.
The leak: which models might come
This is a leak, not officially confirmed. Shortly after the official teaser on November 6, TestingCatalog posted on X that it expects Qwen4-Max, Qwen4-Flash, Qwen4-Plus and Qwen4-27B to arrive soon (post). Alibaba hasn't confirmed these names or published any Qwen4 model ID.
What isn't known yet
When Qwen 4 ships, its final name, model IDs, price, context length, whether the weights will be open, and whether it will be callable over an API on day one. We won't guess.
Use the Qwen3.8 models today
| Model | Model ID | Notes |
|---|---|---|
| Qwen3.8-Max | qwen3.8-max |
The flagship of the Qwen3.8 series, for hard tasks (model page) |
| Qwen3.8-Flash | qwen3.8-flash |
Alibaba's official version of Qwen3.8-Flash-Next, the Qwen4 architecture preview; faster and lighter (model page) |
Both use the OpenAI-compatible API at https://api.agentapiai.com/v1.
curl:
curl https://api.agentapiai.com/v1/chat/completions \
-H "Authorization: Bearer $TEAMO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.8-max",
"messages": [{"role": "user", "content": "Introduce yourself in three sentences."}]
}'
Python (OpenAI SDK):
from openai import OpenAI
client = OpenAI(base_url="https://api.agentapiai.com/v1", api_key="YOUR_TEAMOROUTER_API_KEY")
resp = client.chat.completions.create(
model="qwen3.8-flash",
messages=[{"role": "user", "content": "Rewrite this function in TypeScript: ..."}],
)
print(resp.choices[0].message.content)
Full parameters: API integration docs.
When Qwen 4 ships: if it becomes available on TeamoRouter, change only model in the code above to the ID Alibaba publishes. The base URL and key stay the same.
We'll update this page as soon as Qwen 4 ships
When Qwen 4 is released, we'll update this article with the official model IDs, the official price, and whether it's available on TeamoRouter. You can also vote for it on the Qwen 4 upcoming page. The models live on TeamoRouter are always listed on the models page.
Paying for usage
No subscription: you top up a balance from $5 and pay only for the tokens you use, by card or with USDT, USDC or ETH. The same key also works for Claude, GPT, Gemini and DeepSeek.
FAQ
Is Qwen 4 out?
No. As of November 7, 2026, Alibaba hasn't released Qwen4, and there is no Qwen4 model on the official Qwen account on Hugging Face. The official Qwen account posted a teaser on X on November 6, without a name or a date.
When will Qwen 4 be released?
Alibaba hasn't given a date. The November 6 teaser only says "The becoming begins." We'll update this article when it ships.
What does 北冥有鱼 其名为_ mean?
It's from the Zhuangzi: "In the northern sea there is a fish, its name is Kun." The poster leaves the name blank. Alibaba hasn't explained it or named the new model.
Which Qwen 4 models will there be?
Alibaba hasn't said. According to a leak that isn't officially confirmed, Qwen4-Max, Qwen4-Flash, Qwen4-Plus and Qwen4-27B.
How is Qwen3.8-Flash-Next related to Qwen 4?
Alibaba's model card calls Qwen3.8-Flash-Next an experimental preview of the architecture that will underpin Qwen4. It isn't Qwen 4 itself. Its official version, Qwen3.8-Flash, works on TeamoRouter today.
Can I use Qwen 4 on TeamoRouter?
Not yet, because Alibaba hasn't released it. Qwen3.8-Max and Qwen3.8-Flash work today. We'll update this article with whether Qwen 4 is available once it ships.