---
title: "Qwen3.8 27B API hosted in the EU"
description: "Run Qwen3.8 27B on inference.eu: a dense open model with image and video input, adjustable reasoning and 256K context, OpenAI-compatible, processed in the EU."
canonical: "https://inference.eu/models/qwen3-8-27b/"
language: "en"
---

Reasoning · Qwen (Alibaba Cloud)

# Run Qwen3.8 27B in the EU through an OpenAI-compatible API

Qwen3.8 27B is the compact dense model of Alibaba's Qwen3.8 generation, released in August 2026. It understands images and video, has controllable thinking and a 256K-token context, and targets coding and long agent tasks. On inference.eu you call it with the OpenAI or Anthropic request format, processed in EU data centres and never used for training.

[Contact sales](https://inference.eu/contact/)[Example request](#example)

1.  0127B
2.  02EU data centres
3.  03Apache 2.0

## What Qwen3.8 27B is

Qwen describes Qwen3.8 as the most capable generation of its open models so far, built on the architecture of Qwen3.5. The 27B model is a dense vision-language model with 27 billion parameters and 64 layers, released under the Apache 2.0 licence. It reads images and video, from diagrams and documents to long recordings.

Its native context is 262,144 tokens and can be extended to about one million. Thinking is on by default, can be switched off per request, and its depth is set with `reasoning_effort`. Reasoning from earlier messages is kept by default through preserved thinking. Qwen highlights gains in coding, professional work, research and long-horizon agent tasks over Qwen3.6.

## What it does well, and where it falls short

Qwen3.8 27B offers a lot for its size: strong coding and agent behaviour, image and video understanding, and wide support in common coding harnesses and tools. Because it is dense and only 27 billion parameters, it is easy to serve and quick to respond, which makes it a good default for teams that want one capable mid-sized model.

It does not reach the depth of very large models such as Kimi K3, GLM-5.3 or Qwen's own much larger Qwen3.8 releases on the hardest tasks. High reasoning effort produces long outputs, and Qwen notes that lowering effort in multi-turn agent tasks does not always save time overall. Its training reflects its maker's choices on sensitive topics.

## Calling Qwen3.8 27B through the API

inference.eu uses the OpenAI request format, and the Anthropic Messages format is available too, so SDKs and coding tools work unchanged. Set the base URL to the inference.eu API, use your inference.eu key and set the model to `qwen3.8-27b`.

Set `reasoning_effort` to `low`, `medium` or `xhigh`; `xhigh` is the default. To answer without thinking, pass `enable_thinking: false` in `chat_template_kwargs` through `extra_body`. Preserved thinking is on by default, so pass assistant messages back unchanged in agent loops. Qwen recommends `temperature` 1.0 and `top_p` 0.95 in thinking mode. Send images as `image_url` parts.

## Data residency for Qwen3.8 27B

On inference.eu, Qwen3.8 27B runs on infrastructure in the European Union. Your prompts, images, video, code and the model's responses are processed in EU data centres, are not sent to Alibaba in China, and are never used to train any model.

Every customer receives a data processing agreement under Article 28 of the GDPR and our subprocessor list. Billing is in euros. The Apache 2.0 licence allows commercial use, modification and redistribution without user limits or revenue thresholds, which keeps legal review simple compared with models that carry their own licences.

## Qwen3.8 27B

Maker

Qwen (Alibaba Cloud)

Made in

China

Type

Reasoning

Parameters

27B

Context window

262,144 tokens

Licence

[Apache 2.0](https://huggingface.co/Qwen/Qwen3.8-27B/blob/main/LICENSE)

Released

August 2026

Model ID

qwen3.8-27b

Processing

EU data centres

Training on your data

Never

Source

[huggingface.co](https://huggingface.co/Qwen/Qwen3.8-27B)

qwen3.8-27b · Python

```
import os
from openai import OpenAI

client = OpenAI(base_url="https://api.inference.eu/v1", api_key=os.environ["INFERENCE_EU_API_KEY"])

response = client.chat.completions.create(
    model="qwen3.8-27b",
    messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}],
)
print(response.choices[0].message.content)
```

## What teams use it for

-   1

    ### Default model for a product team

    Use one mid-sized model for chat, coding help, document questions and image analysis across an application.

-   2

    ### Video and screen recordings

    Summarise recorded meetings, product demos or screen captures and answer questions about what happens in them.

-   3

    ### Coding agents at moderate cost

    Run coding agents in common harnesses with adjustable reasoning effort and code processed in the EU.

## Qwen3.8 27B: common questions

The questions buyers, developers and legal teams ask us most. Something missing? Ask us directly.

Compliance depends on how you process data. On inference.eu, Qwen3.8 27B requests are processed in EU data centres, not used for training and covered by a data processing agreement under Article 28 GDPR with a published subprocessor list. Nothing is sent to Alibaba or to China.

The native context is 262,144 tokens, extensible to about one million tokens according to Qwen. Images and video take up part of the context, and reasoning output needs room too, especially at the default \`xhigh\` effort.

Yes. Qwen3.8 27B is released under the Apache 2.0 licence, which allows commercial use, modification and redistribution without fees or user limits. You can build it into paid products without an agreement with Alibaba.

Pass \`reasoning\_effort\` as \`low\`, \`medium\` or \`xhigh\`; the default is \`xhigh\`. Low optimises for speed and cost. Qwen notes that in multi-turn agent tasks lower effort can lead to weaker analysis and more steps, so test what works for your workflow.

Yes. Pass \`enable\_thinking: false\` inside \`chat\_template\_kwargs\` in the request's \`extra\_body\`. Qwen recommends \`temperature\` 0.7 and \`top\_p\` 0.8 for this non-thinking mode, which gives direct answers without a reasoning block and suits simple tasks.

Yes. It is a native vision-language model that understands images and video, from STEM diagrams and documents to hour-long recordings. It answers in text and cannot generate images or video.

Both are dense 27B vision-language models with 256K context. Qwen3.8 27B is newer, with gains in coding, research and agent tasks, plus video understanding and a \`reasoning\_effort\` setting. For new projects it is the better default.

Both are dense multimodal models of similar size with 256K context and Apache 2.0 licences. Qwen3.8 27B adds video input and adjustable reasoning aimed at agentic coding. Gemma 4 31B comes from Google. Compare them on your own prompts.

On inference.eu the open weights run in EU data centres, so your data does not go to Alibaba or to China. The model's training still reflects its maker's choices, including on politically sensitive topics. Test it on your own tasks before relying on it.

Change your OpenAI client's base URL to the inference.eu API, use your inference.eu key and set the model to \`qwen3.8-27b\`. Messages, images, tools and streaming keep working. Add \`reasoning\_effort\` to balance depth against cost.

[Ask a question](https://inference.eu/contact/)

## Related

[All models](https://inference.eu/models/#directory)

-   [Qwen3.6 27BRun Qwen3.6 27B on inference.eu: a dense open model for agentic coding with image input and 256K context, OpenAI-compatible and processed in the EU.Code](https://inference.eu/models/qwen3-6-27b/)
-   [Gemma 4 31BRun Google's Gemma 4 31B on inference.eu: an open multimodal model with 256K context behind an OpenAI-compatible API, processed in EU data centres.Vision](https://inference.eu/models/gemma-4-31b/)
-   [GLM-5.3-FlashRun Z.ai's GLM-5.3-Flash on inference.eu: a multimodal open model with 18B active parameters and 1M context, OpenAI-compatible, processed in the EU.Reasoning](https://inference.eu/models/glm-5-3-flash/)
-   [Qwen3 235B A22B InstructRun Qwen3 235B A22B Instruct (2507) on inference.eu: a 256K-context open model behind an OpenAI-compatible API, processed in EU data centres.Chat](https://inference.eu/models/qwen3-235b-a22b-instruct/)

## Move your AI workloads to EU infrastructure

Tell us which models you use and roughly how much traffic you have. We reply within one working day with pricing and next steps.

-   Reply within one working day
-   Pricing for your volume
-   DPA ready to sign