--- title: "gpt-oss-120b API hosted in the EU" description: "Run OpenAI's open-weight gpt-oss-120b reasoning model on inference.eu: the OpenAI API format, with requests processed in EU data centres, billed in euros." canonical: "https://inference.eu/models/gpt-oss-120b/" language: "en" --- Reasoning · OpenAI # Run gpt-oss-120b in the EU through an OpenAI-compatible API gpt-oss-120b is OpenAI's larger open-weight reasoning model: 117 billion parameters with 5.1 billion active per token, released under Apache 2.0. On inference.eu you keep the OpenAI SDK and request format you already use, but your prompts are processed in EU data centres and never used for training. [Contact sales](https://inference.eu/contact/)[Example request](#example) 1. 01117B · 5.1B 2. 02EU data centres 3. 03Apache 2.0 ## What gpt-oss-120b is gpt-oss-120b is one of two open-weight models OpenAI released in August 2025. It is a mixture-of-experts model with 117 billion parameters, of which 5.1 billion are active for each token, and a context window of 128K tokens. Its expert weights were post-trained in MXFP4 precision, which lets the full model run on a single 80 GB GPU. It is a reasoning model: before answering it works through the problem in a chain of thought, and you choose the reasoning effort per request, from low for quick replies to high for careful analysis. It was trained on OpenAI's harmony response format and supports function calling and structured outputs. It handles text only. ## What it does well, and where it falls short gpt-oss-120b is strong at reasoning, maths, coding and tool use for its active size, and it is a natural first step for teams moving off OpenAI's hosted models: the behaviour and prompt style feel familiar, and structured outputs and function calling work the way OpenAI users expect. The low number of active parameters keeps it efficient. It does not accept images or audio. Its knowledge is older than that of 2026 models such as Qwen3.8 or DeepSeek V4, which beat it on many agentic coding tasks. OpenAI does not publish a list of supported languages on the model card, so test European languages on your own content. High reasoning effort can produce many output tokens. ## Calling gpt-oss-120b through the API If you use the OpenAI SDK today, switching is two lines: set the base URL to the inference.eu API and use your inference.eu key. Then set the model to `gpt-oss-120b`. The harmony format is handled on our side, so you send ordinary chat messages and receive ordinary responses. Set `reasoning_effort` to `low`, `medium` or `high` to trade speed against depth. Tool calls, JSON schema outputs and streaming work as usual. The model's chain of thought is meant for debugging rather than for end users, so show only the final answer in your product. The Anthropic Messages format is available as well. ## OpenAI's model, processed in the EU Using gpt-oss-120b on inference.eu is not the same as using OpenAI's own API. The open weights run on infrastructure in the European Union, your prompts and the responses are processed in EU data centres, nothing is sent to OpenAI, and nothing is used to train any model. Every customer receives a data processing agreement under Article 28 of the GDPR and our list of subprocessors, so there is no transfer of prompts to the United States to assess. Billing is in euros. The Apache 2.0 licence allows commercial use without user limits; OpenAI also publishes a short usage policy for the gpt-oss models. ## gpt-oss-120b Maker OpenAI Made in United States Type Reasoning Parameters 117B Active parameters 5.1B Context window 131,072 tokens Licence [Apache 2.0](https://huggingface.co/openai/gpt-oss-120b/blob/main/LICENSE) Released August 2025 Model ID gpt-oss-120b Processing EU data centres Training on your data Never Source [huggingface.co](https://huggingface.co/openai/gpt-oss-120b) gpt-oss-120b · 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="gpt-oss-120b", messages=[{"role": "user", "content": "Summarise this contract in five bullet points."}], ) print(response.choices[0].message.content) ``` ## What teams use it for - 1 ### Leaving a US AI provider Keep your OpenAI SDK code and prompt style, and move a workload to an EU-hosted open model with a two-line change. - 2 ### Analytical assistants Answer questions that need several steps of reasoning, such as checking contracts against rules or explaining financial figures. - 3 ### Agents with structured outputs Run agents that call tools and return strict JSON schemas, with adjustable reasoning effort per step. ## gpt-oss-120b: 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, gpt-oss-120b 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. Your prompts do not go to OpenAI or to the United States. No. gpt-oss-120b is a separate open-weight model that OpenAI released for anyone to run. It is not the model behind ChatGPT. On inference.eu it runs on our EU infrastructure, independent of OpenAI's own services. gpt-oss-120b supports a context window of 128K tokens, which is 131,072. Reasoning tokens count towards the output, so when you use high reasoning effort leave enough room for both the reasoning and the final answer. Yes. gpt-oss-120b is released under the Apache 2.0 licence, which allows commercial use, modification and redistribution without user limits. OpenAI also asks users to follow a short gpt-oss usage policy, which mainly requires lawful use. Pass \`reasoning\_effort\` with \`low\`, \`medium\` or \`high\`. Low gives fast answers for general dialogue, medium balances speed and detail, and high produces deep analysis at the cost of more output tokens and time. gpt-oss-120b has 117 billion parameters with 5.1 billion active; gpt-oss-20b has 21 billion with 3.6 billion active. The 120b model is stronger on harder reasoning and coding. The 20b model is lighter and suits simpler or high-volume tasks. No. gpt-oss-120b is a text-only model. For images, scanned documents or screenshots, use a multimodal model such as Gemma 4, Qwen3.8 27B or Mistral Small 4, or DeepSeek-OCR 2 for document text. Yes. The model supports function calling and structured outputs natively. Define tools in the standard \`tools\` parameter and use JSON schema response formats as you would with OpenAI's hosted models. Both are mixture-of-experts reasoning models with few active parameters. DeepSeek V4 Flash is newer and larger, with a one-million-token context and stronger agentic coding. gpt-oss-120b has a 128K context and follows OpenAI's familiar response style. No. OpenAI published the weights but is not involved when you call gpt-oss-120b on inference.eu. Requests are processed in EU data centres, used only to produce your response and never used for training. [Ask a question](https://inference.eu/contact/) ## Related [All models](https://inference.eu/models/#directory) - [gpt-oss-20bRun OpenAI's compact open-weight gpt-oss-20b reasoning model on inference.eu: the OpenAI API format, processed in EU data centres, billed in euros.Reasoning](https://inference.eu/models/gpt-oss-20b/) - [DeepSeek V4 FlashRun DeepSeek V4 Flash (0731) on inference.eu: a 1M-token open reasoning model behind an OpenAI-compatible API, processed in EU data centres, not in China.Reasoning](https://inference.eu/models/deepseek-v4-flash/) - [Qwen3.8 27BRun 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.Reasoning](https://inference.eu/models/qwen3-8-27b/) - [Magistral Small 1.2Run Magistral Small 1.2, Mistral AI's open 24B reasoning model, on inference.eu through an OpenAI-compatible API with processing in EU data centres.Reasoning](https://inference.eu/models/magistral-small/) ## 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