Qwen 3.7 release date: May and June 2026, and it is callable today.

Quick answer: the verified two-step timeline.

Qwen 3.7's release happened tier by tier. Qwen3.7-Max, the proprietary text flagship, surfaced first as an unlabeled preview on a public arena leaderboard around May 14, 2026, went live on Alibaba Cloud Model Studio's API on May 19, and was formally announced at the Alibaba Cloud Summit on May 20, per launch coverage. Qwen3.7-Plus, the multimodal volume tier, followed on June 1, 2026 with general availability on the Bailian platform after roughly two and a half weeks of public preview signal (MarkTechPost dated its launch write-up June 2, 2026). So the release-date question has a closed answer with a footnote: the family is out, both ids are stable, and which date you cite depends on which tier you mean. The sections below cover the family history, what shipped in each tier, and what remains unconfirmed about where the line goes next.

The Qwen 3.x family history: three generations in four months.

Alibaba's 2026 cadence through the 3.x line has been roughly a generation every two months, and the family tree explains most of the confusion in search results. Qwen 3.5 arrived on February 16, 2026, alongside a 3.5 Plus tier, with Alibaba claiming wins over US rivals on speed and cost metrics. The 3.6 generation rolled out through April 2026 as a series rather than a single day: open-weight models at multiple sizes (a 35B model on April 16 and a 27B on April 22 per release trackers), a Max preview on April 20, and a 3.6 Plus tier covered by guides that same month. Qwen 3.7 then shipped in the May-to-June two-step described above. Note the pattern, because it repeats: Alibaba releases generations as staggered tiers, previews on public arenas first, flagship announcement second, volume tiers after. A "release date" for any Qwen generation is genuinely several dates, which is why sources disagree without being wrong.

What shipped in each tier: verified vs launch-reported.

One pricing note without numbers, since the table below carries those: the launch made Plus the family's price story, at roughly one-sixth of Max's input rate per launch coverage's math, undercutting even the previous generation's Plus tier. That inversion, newer and lower-priced, is unusual enough that this catalog's qwen3.6-plus entry costing more than qwen3.7-plus routinely surprises people; it matches the vendor's own tiering.

• Verified by consistent coverage: Max is text-only and proprietary; Plus adds image and video understanding on the same 3.7 backbone; both carry a 1M-token context window.
• Verified by the vendor's launch framing: Max targets complex code generation, multi-step agent workflows, and long-horizon execution; Plus targets multimodal and high-volume agent pipelines at a much lower rate.
• Launch-reported, treat as snapshots: third-party comparisons put the Max-over-Plus gap widest on agentic benchmarks (near 8.5 points on one terminal-agent evaluation) and small on knowledge tasks (near 1 point), with Max about 10% faster on text.
• Launch-reported spec-sheet detail: output ceilings of roughly 64K tokens for Max and 32K for Plus. Verify on your serving route if long single generations matter to your workload.

What is not confirmed.

No Qwen 3.8 or Qwen 4 has been announced, and Alibaba has published no roadmap dates beyond the June Plus GA. The 3.x cadence, February, April, May-June, invites extrapolation toward a late-summer 2026 generation, but a pattern is not a date, and this page will not turn one into one. Two softer uncertainties are worth carrying. Third-party benchmark deltas between Max and Plus are launch-week measurements, and both tiers have received platform-side updates since; re-verify before quoting them in a decision document. And multimodal serving is tier- and route-dependent: Plus's vision capability is an upstream property, and the catalog entries here serve text chat completions, so test image input on your actual route before building on it. The practical hedge is the standard one: integrate against exact ids behind an endpoint serving multiple families, and the next generation, whenever it lands, becomes an A/B rather than a migration.

How to call Qwen 3.7 today.

Both released tiers are addressable now through any OpenAI-compatible client: base URL https://api.apisrouter.com/v1, model qwen3.7-max or qwen3.7-plus, one key that also covers qwen3.6-plus for pinned legacy workloads. No Alibaba Cloud account or regional activation is required through the gateway, and the starting credit covers the first requests while you evaluate. All three Qwen ids serve a 1M-token context here, matching the vendor's published window for the 3.7 family.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["APISROUTER_API_KEY"],
    base_url="https://api.apisrouter.com/v1",
)

resp = client.chat.completions.create(
    model="qwen3.7-max",   # announced May 20, 2026; Plus followed June 1
    messages=[{"role": "user", "content": "What changed between Qwen 3.6 and Qwen 3.7?"}],
    max_tokens=2000,
)
print(resp.choices[0].message.content)

Pay-as-you-go · transparent per-model pricing

Selected models are priced below official list prices. Exact input, output, cache, and per-request prices are shown for each model.

Why Qwen release dates keep confusing people.

Four patterns generate the noise, and the 3.7 launch exhibited all of them. Arena previews before announcements: Max was measurable on a public leaderboard days before Alibaba said its name, so "it is already out" and "it is not announced" were simultaneously true. Staggered tiers: Max and Plus shipped twelve days apart, and coverage of each reads as the generation's release. API-before-press: Model Studio served the model a day ahead of the Summit announcement. And generation naming: 3.6 itself rolled out as multiple models across a week, so family dates blur even before 3.7 enters the search results. The reliable check beats all of it: list the models your endpoint serves and look for the exact ids. If qwen3.7-max and qwen3.7-plus appear in /v1/models, the generation is released for you; everything else is press-cycle detail.

curl -s https://api.apisrouter.com/v1/models \
  -H "Authorization: Bearer $APISROUTER_API_KEY" | grep qwen3.7

FAQ