What happened
QwenCloud describes Token Plan Individual as a subscription for individual developers, with a shared Credits balance covering model calls and Harness tools. Its supported-model table now includes qwen4-max-preview, with reasoning, visual understanding, and text generation listed as capabilities. The documentation also identifies qwen4.0-max-preview as an accepted alias; bills uniformly show that alias.
The preview period runs from 10:00 on November 11, 2026, through 24:00 on November 25, 2026, UTC+8. QwenCloud says the model will be retired after the preview ends and says it will make no changes during the period. The promotional rate consumes as little as 20% of the standard Credits rate, described as five times the usage. Calls made from 22:00 to 08:00 UTC+8 receive a further 60% discount, leaving usage at 8% of the standard rate. QwenCloud reserves the right to adjust promotional details or validity periods based on operational needs, and directs users to its latest page or notifications for current terms.
The model is offered through the Token Plan Individual service, rather than as a standalone price-per-token API listing in this documentation. The plan works with mainstream programming and agent tools, including Claude Code, Cursor, Qwen Code, Qoder, and OpenClaw. Plans range from Lite at $6 per month for 20,000 Credits to Pro at $43 per month for 200,000 Credits; the page also lists limited-time new-user prices for Lite and Essential, and says those promotions end November 30, 2026. Credit consumption varies by model type, token usage, thinking mode, and tool calls, so QwenCloud advises checking actual usage in its Model usage view.
The subscription terms matter for teams evaluating the preview. Individual plans are for personal interactive use inside programming and agent tools; the documentation prohibits automation scripts, custom application backends, and non-interactive batch calls. It specifies Singapore as the only available region and Global as the deployment mode, with prompts and outputs involving cross-border data transfer. Subscription quotas reset monthly, unused quota does not carry over, and service pauses when the plan quota and any purchased Credit Pack balance are exhausted.
What the archive adds
New Runtime’s earlier coverage on September 5, 2026, described Qwen3.8-Flash-Next as an early architectural preview for the Qwen4 generation. The current availability is a new fact on that broader Qwen4 thread: a named preview model can now be evaluated through a paid developer subscription, but only for a specified window.
Trend signal
This does not yet fit a pattern New Runtime tracks. The announcement establishes a temporary model-access promotion, not a durable change in how agent systems are priced, integrated, or verified.
Why now
QwenCloud has set a fixed preview window and layered a night discount on top of the promotional credit rate. That timing creates a short evaluation opportunity; the page does not explain why the preview begins or ends on those dates.
Why it matters
For a developer already using QwenCloud’s individual plan, the promotion lowers the cost of trying a new model in familiar programming and agent tools. It may help teams assess whether the model’s stated reasoning and visual-understanding capabilities are useful in their own workflows before deciding whether to pursue another access route. The value is especially practical for bounded experiments that fit within the November 11–25 window.
The limitations are equally relevant. This is not documented as a durable production endpoint: QwenCloud says the model will be retired after the preview. The personal plan terms rule out unattended batch evaluation and custom backend use, so teams cannot assume the offer supports every benchmark or deployment workflow. Credits also vary with model behavior and tool calls, which makes the headline discount an imperfect proxy for the cost of a real task. Treat it as a chance for interactive evaluation, not as a stable procurement commitment.
What happens next
- QwenCloud will retire qwen4-max-preview after the stated preview window unless it updates the offer. (3 months, 75%). Watch for: the model disappearing from the supported-model table or a revised access notice.
Limits and uncertainty
QwenCloud’s page states the access dates, discount rates, listed capabilities, and retirement plan, but it does not report benchmark results or measured performance for this preview. Credits are dynamically determined by model type, token use, thinking mode, and tool calls, so actual task costs are not fixed by the promotional percentages alone. The page reserves QwenCloud’s right to modify the promotion. It also does not establish whether the model will be offered after November 25 through another service or under different terms.
Skill to build now
Set up one small, interactive evaluation in the tool you already use: choose a representative coding or visual-understanding task, record the model ID and prompt, and inspect the resulting Credit usage in QwenCloud’s Model usage view. Run it before the preview closes, and avoid batch automation because the Individual plan terms prohibit non-interactive batch calls.
Reader takeaway
Qwen4-Max is temporarily testable through QwenCloud’s individual developer subscription at a sharply discounted Credits rate, with an additional overnight discount. The useful move is to run a small, interactive evaluation during the stated window; do not assume the preview will remain available or that its personal-use plan supports production automation.