Free Qwen3.8 Max and Flash API on AIHubMix: Limits, End Date, and Your First Call

Update (October 27, 2026): The free campaign has ended. GMI Cloud's campaign credits ran out before the planned end around November 13, and qwen3.8-max-free and qwen3.8-flash-free now return 404 model_retired. To keep using Qwen3.8, drop -free from the model ID and call the paid qwen3.8-max or qwen3.8-flash.

Alibaba's flagship Qwen3.8 Max and its fast sibling Qwen3.8 Flash are free to call on AIHubMix right now. The free model IDs are qwen3.8-max-free and qwen3.8-flash-free, and both cost $0 per million tokens for input and output. GMI Cloud is funding the campaign. It went live on October 14, 2026 (Pacific Time) and runs for 30 days, so until around November 13, or until the campaign credits run out, whichever comes first.

The free IDs follow the same rules as every other AIHubMix free model: 10 trial calls on a new account, then daily quotas once you top up $1 or more. You call them with the standard OpenAI SDK pointed at https://aihubmix.com/v1. When the campaign ends, drop -free from the model ID and the same code keeps running on the paid versions.

What's free, and what it costs the rest of the year

Prices are per million tokens. Qwen list prices are QwenCloud's.

Both free IDs carry the standard capabilities of their paid counterparts: a 1M-token context window, up to 131K output tokens, tool calling, and thinking switched on by default. They're served by GMI Cloud rather than Alibaba Cloud, and the model pages for Qwen3.8 Max (free) and Qwen3.8 Flash (free) show the live latency and uptime for that route.

Two models, two jobs

Benchmark figures below are Qwen's own launch numbers, as compiled by DataCamp; nobody has reproduced them independently.

Qwen3.8 Max launched on August 3, 2026. It's a 2.4-trillion-parameter mixture-of-experts model with roughly 95B parameters active per token, and Qwen pitches it at coding and "Cowork" office tasks. Its launch table shows 67.7 on SWE-bench Pro, 74.8 on CoWorkBench and 92.6 on GPQA Diamond. A post-training update (0902) followed on September 2. Artificial Analysis ranks that version 33rd of 217 models on its Intelligence Index and calls it "very verbose": it used 190M output tokens to run the index, against a median of 87M.

Qwen3.8 Flash launched on August 26, 2026, as the hosted version of the open-weight Flash-Next model (125B parameters, 6B active). Qwen's release table puts it at 62.5 on SWE-bench Pro and 73.9 on CoWorkBench: within a point of Max on office work, at under a thirteenth of Max's output price.

The short version: Flash answers in seconds and covers most everyday work; Max is the one to reach for on hard reasoning and multi-file coding. The second post in this series compares them task by task.

How the free quota works

The AIHubMix free models page sets the limits, and they apply to the Qwen3.8 free IDs as they do to the rest of the catalog:

• Before your first top-up: 10 trial calls in total, shared across every free model, with no expiry. After the tenth, the API keeps replying with a note that the trial is used up instead of returning an error.
• After any top-up of $1 or more: your account switches permanently to daily quotas of 100 requests a day, 10 a minute, and 1M tokens a day. They're shared across the whole free catalog and reset daily.
• Past a cap: the API returns 429 until the next minute or day starts.

Three details about the token pool are easy to miss:

1. It's one pool. Tokens you spend on Flash, or on any other free model, come out of the same 1M as Max.
2. Thinking counts. Both models think before they answer unless you turn it off, and Alibaba bills thinking as output tokens. In a test on October 16, a two-sentence answer from qwen3.8-flash-free used 89 output tokens, 36 of them on thinking.
3. The context window is as big as the daily pool. One request that fills Max's 1M-token context would use the whole day's tokens.

What the free tokens are worth

Assume you use the full 1M tokens a day, split 3:1 between input and output, and price them at AIHubMix's paid rates:

At paid rates, a token spent on Max is worth about 14 times a token spent on Flash. The pool is shared, so the free campaign is worth the most when you spend it on the Max calls where quality matters, and use Flash where speed matters.

Your first call

1. Sign up at AIHubMix and create an API key; the quick start walks through it.
2. Top up $1 or more if you want the daily quota rather than 10 trial calls.
3. Export the key as AIHUBMIX_API_KEY and send a request:

curl https://aihubmix.com/v1/chat/completions \
  -H "Authorization: Bearer $AIHUBMIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-flash-free",
    "messages": [{"role": "user", "content": "In two sentences, what is a mixture-of-experts model?"}]
  }'

For Max, stream the response. Max thinks for a while before its first visible token: AIHubMix's own measurement puts the free route's latency at about 55 seconds, against about 5 seconds for Flash. Streaming keeps data flowing while it thinks, and a longer client timeout covers long answers:

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://aihubmix.com/v1",
    api_key=os.environ["AIHUBMIX_API_KEY"],
    timeout=600,
)

stream = client.chat.completions.create(
    model="qwen3.8-max-free",
    messages=[{"role": "user", "content": "Find the bug in this function:\n\ndef mean(xs):\n    return sum(xs) / len(xs) - 1"}],
    stream=True,
)

for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

The thinking arrives in a separate reasoning_content field, so this loop prints only the answer. For short, simple requests, you can switch thinking off with extra_body={"enable_thinking": False}, which is Alibaba's documented switch for Qwen3.8; the third post in this series shows how to confirm it took effect.

When the free window closes

The -free IDs stop being free around November 13, 2026 (Pacific Time), or earlier if GMI Cloud's campaign credits run out first. Two ways to make that a non-event:

• Change one string. Swap qwen3.8-max-free for qwen3.8-max, or qwen3.8-flash-free for qwen3.8-flash. Endpoint, SDK and parameters stay the same, and the AIHubMix prices are below Qwen's own list prices.
• Set up fallback now. AIHubMix can fall back from a free model to a paid one at the key level when the free model hits a quota or rate limit, with no code change. The third post walks through the setup.

FAQ

Is Qwen3.8 Max really free on AIHubMix?
Yes. qwen3.8-max-free costs $0 per million input and output tokens during GMI Cloud's campaign. The normal AIHubMix free-model quotas apply.

When does the free Qwen3.8 offer end?
30 days after its October 14, 2026 launch (Pacific Time), so around November 13, or earlier if the campaign credits run out.

Do I need a credit card to try it?
No. Every new account gets 10 trial calls. A top-up of $1 or more switches the account to daily quotas of 100 requests and 1M tokens.

Is the free version the same model as the paid one?
The free IDs carry the same standard capabilities: 1M context, up to 131K output, tool calling and thinking. They run on GMI Cloud instead of Alibaba Cloud, so latency and throughput differ; the model pages show both.

Do thinking tokens count against the free quota?
Yes. Thinking is output, and every token counts toward the shared 1M-token daily pool. Turn thinking off for simple requests to make the pool last longer.

Can I use the free models in coding tools?
Yes, in any tool that accepts an OpenAI-compatible base URL and a custom model ID, such as OpenCode, Cline, Kilo Code or Qwen Code. Agent loops resend context on every turn, so they drain the daily pool quickly.

What happens to my code when the offer ends?
Change the model ID to qwen3.8-max or qwen3.8-flash and nothing else. Or configure key-level fallback now so the switch happens on its own.

Keep reading: the Qwen3.8 Free series

• The pool is shared, so it pays to know which model deserves your tokens: Qwen3.8 Max Free vs Qwen3.8 Flash Free: Is Flash Enough?
• Once you're calling it every day, here's what goes wrong and the one setting that covers most of it: Qwen3.8 Free on AIHubMix: Fix 429s, "Rate Limited by Provider", and a Drained Token Pool

Sources

• Qwen3.8 Max (free) model page (AIHubMix)
• Free AI models and quotas (AIHubMix)
• Use deep thinking models via API (Alibaba Cloud Model Studio)
• Qwen3.8-Max: Features, Benchmarks, and Pricing (DataCamp)
• Qwen3.8 Max (0902): Intelligence, Performance & Price Analysis (Artificial Analysis)