Free Qwen API: Alibaba Cloud quotas and the price after them

The query "free Qwen API" usually means two different things: "I want to try it without paying" and "I never want to pay". Alibaba Cloud has an answer to the first in the form of a new-user quota. For the second there is no honest answer: when the quota runs out, charges begin. Below are the quota terms according to the Alibaba Cloud Model Studio documentation (the page was updated on 16 October 2026) and a calculation of what Qwen 3.8 Flash costs after it.

Free Qwen API: what Alibaba Cloud gives

According to the documentation, when you first activate Model Studio in the Singapore region, the platform automatically grants free quotas for various models. These are the conditions worth reading before you count on "free":

• Only certain models. The quota applies to models in the Singapore region whose service deployment scope is International. Models in other regions or with a different scope are not eligible.
• A time limit. The quota is valid for 90 days. That term is stated for those who activate the service for the first time from 8 September. Any unused remainder expires when the period ends, even if you never touched it.
• A separate quota per model. Each model has its own limit, described in the docs as "typically 1,000,000 tokens". A remainder cannot be moved from one model to another. A model version with a date in its name and one without are treated as two different models.
• Input and output are counted together. The console shows a single remaining total, and each request deducts both input and output tokens.
• What is not covered. The quota covers real-time invocations. Batch calls, fine-tuning and deploying your own models are excluded.
• Re-registering gives nothing. A new account does not bring an extra quota.
• The quota is shared by the account and its RAM users.

The figures and dates in this list come from a single documentation page. They have changed before and may change again, so open the current "Free quota for new users" page before you sign up.

What happens when the quota runs out

The most important point in the docs: by default, once the quota is used up you automatically move to pay-as-you-go billing. The money is charged to your Alibaba Cloud account, and if a payment is overdue the service may become unavailable.

To avoid this Alibaba has a Free Quota Only mode: when it is on, the service stops responding once the quota is exhausted and returns the error AllocationQuota.FreeTierOnly, with no charges. The mode is off by default and has to be enabled per model in the console (or in batch). The switch does not take effect instantly, so do not start a heavy workload right after turning it on.

Two more things that make "free" unexpectedly paid:

• IDE plugins burn tokens quickly. The docs warn explicitly: with Cline and similar plugins each request carries the code context and the conversation history, so one exchange can cost thousands of tokens.
• The console refreshes the remainder once a minute and the page has to be refreshed by hand. You can see a remainder that is already gone.

Alibaba sends notifications when the remainder drops to 20% and when the quota is exhausted. It is also worth setting up a spending alert here.

What Qwen 3.8 Flash costs without a quota

Suppose you have spent the quota and want to continue. Zukko prices as of 18 October 2026, $ per 1M tokens:

To compare the size of a "free" million tokens with money, take a hypothetical split (an assumption, not a market average) of 80% input and 20% output: 800K input and 200K output tokens.

• qwen3.8-flash at Zukko: 0.8 × 0.038 + 0.2 × 0.075 = 0.0304 + 0.015 = $0.0454.
• qwen3.8-flash at official prices: 0.8 × 0.15 + 0.2 × 0.47 = 0.12 + 0.094 = $0.214.

So a free million tokens on one model is worth about 4.5 cents at the gateway. This is not a criticism of the quota, just a sense of scale: if you are trying a model, the quota covers the whole task; if you are building a product on it, it may not last even one load test.

The minimum top-up at Zukko is $1. At the same 80/20 split that is roughly 1 / 0.0454 ≈ 22M tokens on qwen3.8-flash. The figure is approximate: your real input/output ratio will differ, so calculate from your own usage logs. How a request price is built up is covered in the article how request price is calculated↗.

What Zukko offers for free access

Zukko has no free mode: it is a pay-per-token gateway with a dollar balance and top-ups from $1. You can top up via SBP (the payment page adds an 8.5% fee on top), cryptocurrency or a MasterCard. To avoid a surprise bill, each key can have a daily and a monthly limit in dollars: when a limit is reached the key returns 429 (key_daily_limit_reached or key_monthly_limit_reached) and no more money is spent.

A practical plan for a beginner:

1. If you need to test an idea on Qwen and are ready to register in the Alibaba cloud, take their quota, enable Free Quota Only and watch the expiry date.
2. If the quota is gone or registration does not suit you, create a key in the dashboard↗, set a limit of $1–2 and call qwen3.8-flash at https://api.zukko.pro/v1.
3. If you want to compare Qwen with other models, do not change the key or the address, only the model name. Current models and prices are in the models reference↗ and in GET /v1/models.

There is a separate overview of other cheap and partly free API options: free neural networks via API↗. Ways to run large volumes cheaply are covered in the article on the cheapest models for bulk tasks↗.

FAQ

Is there a free Qwen API key with no payment details attached?

At Alibaba the quota is granted when you activate the service, and you must complete your account information first. Which data and payment methods are required, check in the Alibaba Cloud documentation and console, since they can change.

Can I get the quota again with a new account?

According to the docs, no: re-registering does not grant an additional quota, and a remainder expires at the end of the term without being extended.

What happens if I exceed the quota without enabling Free Quota Only?

Usage beyond the quota is charged at the console prices on a pay-as-you-go basis. This is normal behaviour, and per the docs such charges are not refunded.

Which Qwen model should I take after the quota?

For simple bulk tasks start with qwen3.8-flash, for harder ones try qwen3.8-max. Their strength index is 40 and 45, and both have a 1M context. Decide by results on your own tasks and by the price in the calculation above.