Trust and compliance · 2026-10-14
Free Qwen in Europe: where requests go and what is stored
heabsy/qwen3.8-27b-heabsy-free is free until 2 November 2026. According to Heabsy, it runs in the EU/EEA, prompts are not stored and are not used for training.
Short answer: there is a free Qwen row processed in the EU/EEA
LLMTR (llmtr.com) is a gateway that gives access to language models hosted in Türkiye and at global providers through one OpenAI-compatible API. You can call Qwen3.8 27B there for free with the model id heabsy/qwen3.8-27b-heabsy-free at https://llmtr.com/v1 until 2 November 2026, 00:00 (TRT). During this period the price is $0.00 per 1M input tokens and $0.00 per 1M output tokens.
Heabsy serves this row. According to Heabsy, Qwen3.8 27B runs in EU/EEA data centres, prompts and responses are not stored, and they are not used to train models. Before you call the row, your account needs a data-transfer consent for Heabsy. Use is limited per account per day; the context window is 262,144 tokens and max_tokens can go up to 32,768.
Where do your requests go when you use a free Qwen?
It depends on which model id you call. The LLMTR catalog carries Qwen3.8 27B in five rows from four providers; two of them are Heabsy's. Your request reaches LLMTR first, then that row's provider; the answer comes back the same way. So there is no single answer to whether you can trust a free Qwen: the model name is the same, but where the request is processed, who stores what and which terms apply change from row to row.
Two fields on every model page answer this. The region badge (EU or Türkiye) shows where the model is hosted. The data note passes on the provider's statement about retention and training, attributed to the provider; on free rows it also mentions the daily quota. Read both before you pick a row.
What Heabsy states: EU/EEA, zero data retention, no training
According to Heabsy, Qwen3.8 27B runs in EU/EEA data centres. Heabsy states that it does not store prompts or responses, which it calls zero data retention, and that it does not use them to train models. The data note on the free row's card passes on the same statement and adds that, under the provider's terms, prompt and response content is not stored in request logs.
Heabsy offers a data processing agreement (DPA), and the Heabsy consent card on LLMTR links to it. For teams with GDPR obligations, what binds is not this summary but the text of the privacy policy and the DPA. Base your assessment of processing location, retention and sub-processors on reading those two documents.
What happens on the LLMTR side: consent, transferred content and retention
Heabsy models cannot be called until your account has granted a data-transfer consent for Heabsy. Without it, every request is refused with 403 provider_consent_required and does not reach Heabsy. You grant the consent under Dashboard > Settings > Data-transfer consents and can withdraw it from the same place at any time. The consent is account-level and applies to every API key on the account.
The consent is per provider: a consent you grant for GreenPT or TensorX does not cover Heabsy rows, and a Heabsy consent does not cover theirs. The other two Qwen3.8 27B rows in the catalog, llmtr/qwen3-8-27b and qwen/qwen3.8-27b-free, do not ask for such a consent.
Only the content of your request is transferred to Heabsy, so that it can produce the answer; account details such as your name and e-mail address are not transferred. LLMTR's usage records do not contain prompt or response bodies; per request they hold metadata such as the model id, token counts, HTTP status and USD cost.
Qwen3.8 27B options in the LLMTR catalog
The table below puts the five rows offered under the Qwen3.8 27B name side by side. Prices are per 1M tokens. The Heabsy and TensorX rows each need their own data-transfer consent; the llmtr/ and qwen/ rows do not.
You can open the same list from a model page. The +N providers button next to the provider badge opens a menu titled "Same model, other providers" that lists each row with the provider name, a region label (EU or Türkiye) and, on free rows, a "free" note. When the free period ends, heabsy/qwen3.8-27b-heabsy-free drops off this menu.
| Model id | Provider | Price (input / output) | Context | Hosting | Data note |
|---|---|---|---|---|---|
| heabsy/qwen3.8-27b-heabsy-free | Heabsy | $0.00 / $0.00 (until 2 November 2026, 00:00 TRT) | 262,144 | EU/EEA (according to Heabsy) | According to the provider, processed in the EU/EEA, inputs not used for training, prompt and response content not stored in request logs; daily quota |
| heabsy/qwen3.8-27b-heabsy | Heabsy | $0.40 / $3.00 (cached input $0.05) | 262,144 | EU/EEA (according to Heabsy) | According to the provider, processed in the EU/EEA, inputs not used for training |
| tensorx/qwen3.8-27b-tensorx | TensorX | $0.40 / $2.40 (cached input $0.10) | 262,144 | EU | According to the provider, processed in the EU, inputs not used for training |
| llmtr/qwen3-8-27b | LLMTR | $5.00 / $7.00 (cached input $1.25) | 262,144 | Türkiye | None |
| qwen/qwen3.8-27b-free | Qwen | $0.00 / $0.00 | 131,042 | Not stated | Prompts and completions may be logged by the provider and used to improve the model; processed on third-party infrastructure abroad; daily quota |
Which row to pick, and a first request
The choice depends on where your data has to be processed and on the card's data note. If you need EU/EEA processing and no retention of prompts, Heabsy's statement covers both: heabsy/qwen3.8-27b-heabsy-free today, and the same provider's paid row heabsy/qwen3.8-27b-heabsy after 2 November 2026. According to its card, tensorx/qwen3.8-27b-tensorx also processes in the EU and does not use inputs for training; for its retention terms, read TensorX's own documents. If hosting has to be in Türkiye, use llmtr/qwen3-8-27b.
For short, non-sensitive work and prototypes, qwen/qwen3.8-27b-free is also free, with a smaller context; its card says prompts and completions may be logged by the provider and used to improve the model. If you send personal or sensitive data, pick a row whose data note fits your team's own obligations; that decision is yours.
The free period ends on 2 November 2026, 00:00 (TRT, UTC+3). After that, requests to heabsy/qwen3.8-27b-heabsy-free return HTTP 410 model_retired, and the error names heabsy/qwen3.8-27b-heabsy, the same model from the same provider at a metered price. When the daily limit is reached, the API returns HTTP 429 with a Retry-After header; wait for that time, then send the request again. On an account that has not topped up yet, a request can return HTTP 402 after some use, with details.reason set to account_verification_required. Automated-use protection is asking for a one-time account verification: top up once from Dashboard > Billing and retry after a minute; the model is not charged to your balance.
The curl request below removes a name and an e-mail address from a customer message. The model reasons by default; if the answer comes back with finish_reason: length, raise max_tokens, up to 32,768. reasoning_effort is not supported by this provider; a request that sends it is refused with HTTP 400.
curl request to heabsy/qwen3.8-27b-heabsy-free that removes a name and an e-mail address from a customer message.
curl https://llmtr.com/v1/chat/completions \
-H "Authorization: Bearer $LLMTR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "heabsy/qwen3.8-27b-heabsy-free",
"max_tokens": 2048,
"messages": [
{"role": "user", "content": "Remove names and e-mail addresses from this text and return the rest unchanged: Hi, I am Jane Doe (jane@example.com). Order 4417 arrived damaged."}
]
}'Call the free Qwen3.8 27B on a row processed in Europe
Grant the Heabsy data-transfer consent, create an API key and send the first request to heabsy/qwen3.8-27b-heabsy-free before 2 November 2026, 00:00 (TRT).
- Data-transfer consent. In your llmtr.com account, grant the Heabsy consent under Dashboard > Settings > Data-transfer consents. Without it, the request is refused with 403 provider_consent_required.
- API key. Create a key under Dashboard > API Keys. The raw key is shown once; store it as the LLMTR_API_KEY environment variable.
- First request. Send the first request to https://llmtr.com/v1 with any OpenAI-compatible SDK or curl, using heabsy/qwen3.8-27b-heabsy-free as the model.
Frequently asked questions
Does the free Qwen3.8 27B store my prompts?
According to Heabsy, prompts and responses are not stored and are not used for model training; Heabsy calls this zero data retention. LLMTR's usage records do not contain prompt or response bodies either; they hold metadata such as token counts.
Why does heabsy/qwen3.8-27b-heabsy-free return 403 provider_consent_required?
Your account has no data-transfer consent for Heabsy. Grant it under Dashboard > Settings > Data-transfer consents; a GreenPT or TensorX consent does not cover Heabsy rows.
What happens to the free Qwen row after 2 November 2026?
Requests return HTTP 410 model_retired, and the error names the paid heabsy/qwen3.8-27b-heabsy. To keep using Qwen3.8 27B on a row processed in Europe, switch the model id to that row or to tensorx/qwen3.8-27b-tensorx.