Qwen 4.0: Release Date, 27B, 1M Context and API Access

Qwen 4.0 release date: is it coming soon?

The strongest public signals are a roadmap photo and integration work, rather than a dated general-availability announcement.

In the Qwen4-27B community thread, the shared stage photo visibly says “Qwen4 系列 Coming Soon” and lists Qwen4-Max, Qwen4-Flash & Qwen4-Plus, and Qwen4-27B. This is a useful roadmap clue, although the Reddit post is a community source rather than the original event recording or release notes.

On October 28, 2026, Qwen Code PR #15762 was merged with the title “fix(core): size the qwen4 generation at 1M.” That is a directly inspectable change in the QwenLM repository.

Treat specific dates circulating beside these clues carefully. The million-context Reddit thread initially called a date November 1, then corrected it to October 1 and suggested it was a placeholder. The string qwen4-max-2026-10-01 in a test is not a launch calendar. Our assessment: Qwen4 looks actively prepared, but “coming soon” cannot yet be translated into a reliable release day.

Qwen4-Max, Plus, Flash and 27B: what do the names mean?

The roadmap photo points to a family rather than one universal Qwen 4.0 model. Max suggests a flagship tier, Plus and Flash separate service tiers, and 27B a parameter-size variant. Their final packaging, licenses, rollout order and public identifiers still need release documentation.

Code references also include qwen4-coder-plus and qwen4-omni-120b-think. The PR describes the latter as a model name used in test fixtures. A fixture is development evidence, not proof of a downloadable 120B release or a public endpoint. Likewise, 27B does not tell us the memory needed at a particular quantization or context length.

Developers searching for “Qwen 4,” “Qwen4” and “Qwen 4.0” are following the same anticipated generation here. These spellings should not be treated as interchangeable production API IDs until the provider publishes the exact identifier.

Does Qwen4 have a 1M-token context window?

Qwen Code now has a 1,000,000-token forward default for names beginning with qwen4. Its tests cover Max, Plus, Coder, dated names and provider prefixes. The catalog takes precedence if a model-specific entry later supplies a different limit.

The PR explicitly states that the CLI was tested against a fake endpoint, with no real Qwen4 endpoint called. It also says the generation-wide window comes from a reviewer report and leaves output ceilings unchanged. Consequently, this is strong evidence of client preparation for 1M, rather than an independent measurement of every Qwen4 model’s actual context.

For users, the distinction affects repository agents and long document sessions: a client can budget for a million tokens while an upstream route still enforces a smaller limit. Wait for the final model card and test the selected route before filling that window. No Qwen4 maximum output length is established by this change.

Qwen 4.0 Max leaks and architecture: what is worth following?

Another community investigation links a T-Head PLA pull request titled “feat(gdn): support Qwen4.0-Max prefill heads.” The referenced GitHub API returned Not Found during our check, so we can report the community claim but cannot independently confirm that change or its contents.

That thread also circulates multi-trillion-parameter expectations. We have no verified model card supporting a Qwen4-Max parameter count, activated-parameter count or benchmark score. A kernel shape or a model name alone cannot supply those numbers.

The useful questions at launch are whether long-context attention becomes cheaper, coding agents finish more tasks per unit of cost, and smaller models improve local deployment. Those are evaluation priorities, not published Qwen4 results. No verified head-to-head result currently establishes Qwen4 as better than Qwen3.8, GPT or Claude.

Qwen4 vs Qwen3.8: should developers wait?

Use an available Qwen3.8 route for work that must run today; keep the model name configurable for Qwen4 evaluation. Flash is the inexpensive baseline, while Max is a second candidate for harder prompts. Measure success rate and total billed tokens before selecting either for a workload.

Qwen4 has roadmap and software-support clues; Qwen3.8-Flash and Max already have public prices and purchasable APIMaster routes. The current official pricing table already lists a 1M input tier for both Qwen3.8 models, so “1M” alone will not establish a generational advantage. Route-specific context, tools and multimodal support still need checking on the market card.

Our Qwen3.8-Flash guide and Qwen3.8-Max guide provide background; use this article’s dated table or the live cards for current prices.

Qwen API key: what can you call now?

Check the Flash card or Max card, register, then create an API key. Use https://apimaster.ai/v1 with an available model ID. This minimal text request uses released Qwen3.8-Flash, not a speculative Qwen4 endpoint:

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_APIMASTER_KEY",
    base_url="https://apimaster.ai/v1",
)
response = client.chat.completions.create(
    model="qwen3.8-flash",
    messages=[{"role": "user", "content": "Review this migration plan and identify its main risks."}],
)
print(response.choices[0].message.content)

Keep a small evaluation set covering your coding, multilingual and document tasks. When Qwen4 becomes publicly available, compare completion quality, latency, tool behavior and full-task cost using its exact published ID. Today’s Qwen3.8 routes offer at least 60% off the stated standard comparison rates without waiting for that rollout.

FAQ

Has Qwen 4.0 been released?

At the November 3, 2026 check, we found roadmap and integration evidence, but did not verify a generally available Qwen4 API, public weight release or release date.

Is Qwen4-27B coming?

A community-shared conference slide lists Qwen4-27B under “Coming Soon.” Its download date, license and final specifications still need a published model card.

Is the Qwen4 context window 1M?

Qwen Code merged a 1M family default. The PR used a fake endpoint and explicitly leaves the actual server limits unconfirmed; a future catalog entry can override the default.

What is the Qwen 4.0 API price?

No verified Qwen4 official price or public APIMaster route was found. The available Qwen3.8 prices in this article belong to Qwen3.8, not Qwen4.

Can I use a Qwen API key today?

Yes. Create an APIMaster key and select an available model such as qwen3.8-flash or qwen3.8-max. Qwen4 access will depend on a future public route and its exact supported identifier.

Sources and check date

Checked November 3, 2026: Qwen4 roadmap photo and 27B discussion, merged Qwen Code 1M-context PR, release-date correction, Max development claims, Alibaba Cloud model pricing, and APIMaster public marketplace responses. Roadmap images, code defaults and community speculation are identified separately throughout.