EN ▾
Get API key

qwenapis.comGuide

Qwen 3 API Pricing: Per-Token Costs and Top-Up Tiers

This service charges $0.15 per 1M input tokens and $0.50 per 1M output tokens from prepaid credit. Below are the rates, the top-up bonuses and worked cost examples you can adapt to your own traffic.

Updated:

Balance scale weighing token cubes against coins to illustrate per-token API pricing

The rates in plain numbers

Anyone looking up qwen 3 api pricing should first know what they are reading. This site is an independent provider that runs one open-weight model, tuned to answer without content refusals for lawful adult use, under the model ID uncensored. It is not the official price list of Alibaba or any other vendor, and it does not resell their plans. The numbers below are what this service charges, nothing more. For a plain account of the model name confusion, see the myths versus facts article.

ItemPrice
Input tokens$0.15 per 1M
Output tokens$0.50 per 1M
SubscriptionNone, prepaid credit only
Expiry of creditNever
Failed or refused requestsNot charged
Trial$0.50, one click, one per person, 7 days, no card

Input tokens are everything you send: system prompt, chat history, documents and tool definitions. Output tokens are what the model writes back. Output costs a bit more than three times as much per token as input, so the ratio of the two in your workload decides the bill. A summarizer sends a lot and receives a little. A story writer is the opposite.

Top-ups, bonuses and the tier edges

You pay in advance. A top-up can be any whole dollar amount from $20 to $500, and a bonus is added to larger ones. The bonus is a percentage of the amount you pay, applied on the whole payment, not only on the part above the threshold.

You payBonusCredit you get
$20none$20.00
$49none$49.00
$50+5%$52.50
$99+5%$103.95
$100+10%$110.00
$200+15%$230.00
$500+15%$575.00

Look at the edges. Paying $49 gives $49 of credit, while paying $50 gives $52.50, so one extra dollar returns $3.50 more credit. The same jump happens between $99 and $100: $103.95 becomes $110.00. If your planned amount lands just below a threshold, move up to the threshold.

At the top tier you are effectively paying less per token. With +15%, $200 buys $230 of credit, so the effective input rate is $0.15 / 1.15 = about $0.130 per 1M and the effective output rate is $0.50 / 1.15 = about $0.435 per 1M. Methods: crypto (USDT on TRC20, USDC on Base). The pricing page shows the live options.

Worked cost examples

Abstract rates are hard to feel, so here are four workloads with the arithmetic shown. Each uses exactly $0.15 and $0.50 per 1M tokens, before any bonus.

Support bot, 20,000 conversations a month

Assume each conversation is 1,200 input tokens (system prompt plus the customer message plus a little history) and 350 output tokens.

  • Input: 20,000 x 1,200 = 24,000,000 tokens = 24 x $0.15 = $3.60
  • Output: 20,000 x 350 = 7,000,000 tokens = 7 x $0.50 = $3.50
  • Total: $7.10 a month

Document summaries, 2,000 files a month

Each file is 20,000 tokens in and the summary is 600 tokens out.

  • Input: 2,000 x 20,000 = 40,000,000 tokens = 40 x $0.15 = $6.00
  • Output: 2,000 x 600 = 1,200,000 tokens = 1.2 x $0.50 = $0.60
  • Total: $6.60 a month

Agent runs, 1,000 runs a month

An agent takes 20 steps per run, resending about 30,000 tokens of context at each step and writing 800 tokens per step.

  • Input per run: 20 x 30,000 = 600,000 tokens = $0.09
  • Output per run: 20 x 800 = 16,000 tokens = $0.008
  • Per run: $0.098, so 1,000 runs cost $98.00

Notice that the agent dominates the bill through resent context, not through the words it writes. Trimming tool output from 30,000 to 15,000 tokens per step would roughly halve the input line.

One nearly full window

The context window is 100,000 tokens for prompt plus reply. A request with 99,000 in and 1,000 out costs 99,000 x $0.15 / 1M = $0.01485 plus 1,000 x $0.50 / 1M = $0.0005, which is $0.01535. At that size, 100 requests cost $1.535.

Estimating your own bill

You do not need a spreadsheet. Measure the average input and output tokens of a few real requests, then multiply. The response includes a usage block, so you can read the numbers instead of guessing.

import os
from openai import OpenAI

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

def cost_usd(usage):
    return usage.prompt_tokens * 0.15 / 1e6 + usage.completion_tokens * 0.50 / 1e6

r = client.chat.completions.create(
    model="uncensored",
    messages=[{"role": "user", "content": "Summarize the plot of a heist film in 3 sentences."}],
)
print(r.usage.prompt_tokens, r.usage.completion_tokens, round(cost_usd(r.usage), 6))

Run that on 20 representative prompts and average the result. Then multiply by monthly volume. Remember that history grows: a chat that resends all earlier turns costs more per message the longer it runs. If your product has long sessions, measure the late turns, not only the early ones.

You can also read the remaining credit from the X-Balance-USD header on any response, which lets you build a low-balance alert without polling a separate endpoint. Because failed or refused requests are not charged, retries after an error do not cost anything.

Comparing against other hosted options

Other providers publish their own per-token prices, and they change. This article does not quote them, because a stale number is worse than none. Check each provider's current pricing page and compare on the same terms. The checklist below covers what moves a real bill more than the headline rate does.

Question to askThis service
Separate input and output prices?Yes, $0.15 and $0.50 per 1M
Subscription or minimum spend?No subscription, minimum top-up $20
Does credit expire?No
Are failed requests billed?No
Rate limits on a paid key?300 requests per minute, 8 concurrent
Context window?100,000 tokens, prompt plus completion
Prompts used for training?No

Be honest about fit, too. If you need image generation, embeddings, fine-tuning or a specific named model from a specific vendor, this service does not offer those, and a vendor's own API is the right place to look. If you need a text model that answers lawful adult requests without refusals, behind an OpenAI-style endpoint that you can try for $0.50 first, it is built for that.

Keeping spend predictable

Three habits keep a per-token bill from surprising you. First, set max_tokens when a feature has a natural length, such as a one-paragraph summary. It is optional on this API, and without it a completion can use the rest of the window, so a runaway answer could cost more than you planned. Output tokens are the more expensive kind at $0.50 per 1M.

Second, trim history. A chat that resends every earlier message pays for the whole transcript on each turn, so a 40-turn conversation costs far more than 40 single turns. Summarize old turns into a few lines, or keep only the last several exchanges plus a short note of facts the model must remember. Third, watch the balance. The X-Balance-USD header lets you raise an alert when credit falls below the amount you spend in a week, and you can top up before requests start failing.

Because credit does not expire and there is no subscription, there is no penalty for buying less at first. Start at $20, measure a week of real use, and then choose a bigger top-up if the bonus tiers suit your volume. Moving to $50 is worth it once you know you will spend it, since the +5% bonus applies from that amount.

Starting without spending much

New users can claim $0.50 of trial credit with one click on the API key page after signing up with Google or with an email and password. No card is needed, it is one per person, and it lasts 7 days. At the support-bot numbers above, one exchange of 1,200 in and 350 out costs $0.00018 + $0.000175 = $0.000355, so $0.50 covers about 1,400 of them. That is plenty to judge answer quality.

A trial key allows 2 concurrent requests and 60 per minute. A full key allows 8 and 300. Beyond the trial, the first real top-up is $20, and the quick-start documentation and the key guide take you from sign-up to a working request in a few minutes.

Questions and answers

How much does this API cost per token?

Input costs $0.15 per 1M tokens and output costs $0.50 per 1M tokens, billed from prepaid credit with no subscription.

Is this the official Qwen 3 price list?

No. It is an independent provider with its own model ID, uncensored, and its own rates. For any vendor's official prices, check that vendor's current pricing page.

What is the smallest top-up and are there bonuses?

The smallest top-up is $20 and the largest is $500, in whole dollars. Bonuses are +5% from $50, +10% from $100 and +15% from $200.

Does my credit expire?

No, prepaid credit never expires. Only the $0.50 trial credit has a time limit of 7 days.

Am I charged when a request fails?

No. Failed or refused requests are not charged, so retrying after an error does not add to your bill.

Your key is one form away

Create an account, copy the key, change the base URL. That is the whole setup.

Get API key