OpenAIClosed weights
GPT-5 mini.
GPT-5 mini is the compact edition of GPT-5 for lighter reasoning tasks, with the same instruction following at lower latency and lower cost. It is the successor to o4-mini.
base_url="https://api.noirdoc.de/v1" model="gpt-5-mini"
- Data centres in the EU
- § 203 confidentiality
- Zero retention
- List prices
Details.
Where the model runs, which rules apply and what it can do.
- Data centres in the EU Azure Contracting party: Microsoft Ireland Operations Limited · based in Dublin, Ireland
- § 203 German Criminal Code (StGB)
- Confidentiality agreed with Azure
- Weights
- Closed weights
- Features
- Tools, Vision, Reasoning and JSON
Providers.
We offer GPT-5 mini through these providers.
Provider § 203 model=
€ · 1M tokens Input Cache Read Cache Write Output
More models.
OpenAI develops the GPT series, Whisper and TTS. Alongside its proprietary models, OpenAI also publishes open weights with GPT-OSS.
All OpenAI models → Model
€ · 1M tokens Input Cache Read Cache Write Output
GPT-6.1 Sol Tools Vision Reasoning JSON GPT-6 model for agentic coding and computer use — also for document-heavy work and multi-step business workflows. 
Azure · EU Azure · EU Azure · EU Azure · EU Azure · EU Azure · EU Azure · EU Azure · EU 
AKI.IO · DE Azure · EU
Azure · EU Input 2.16 Cache Read 0.11 Cache Write 2.70 Output 10.78 from 272K input tokens from 272K Input 4.23 Cache Read 0.22 Cache Write 2.70 Output 15.84
GPT-6 Sol Tools JSON Cost-efficient top tier of the GPT-6 family — for agentic coding, business automation and long software engineering tasks. Azure · EU Input 2.16 Cache Read 0.22 Cache Write 2.70 Output 10.78 from 272K input tokens from 272K Input 4.25 Cache Read 0.44 Cache Write 5.28 Output 15.85
GPT-6 Luna Tools Vision Reasoning JSON Fast and affordable GPT-6 model — for high volumes, chat, classification and light agentic workflows. Azure · EU Input 0.11 Cache Read 0.011 Cache Write 0.135 Output 0.54 from 272K input tokens from 272K Input 0.212 Cache Read 0.022 Cache Write 0.135 Output 0.792
GPT-5.6 Sol Tools Vision Reasoning JSON Most capable model of the GPT-5.6 family — extended reasoning, agentic workflows and code for the most demanding workloads. Azure · EU Input 3.95 Cache Read 0.395 Cache Write 4.94 Output 19.75 from 272K input tokens from 272K Input 7.76 Cache Read 0.776 Cache Write 9.70 Output 29.06
GPT-5.6 Terra Tools Vision Reasoning JSON Balanced model of the 5.6 family — GPT-5.5-level performance at significantly lower cost. Azure · EU Input 1.98 Cache Read 0.198 Cache Write 2.47 Output 11.85 from 272K input tokens from 272K Input 3.90 Cache Read 0.39 Cache Write 4.88 Output 17.46
GPT-5.6 Luna Tools Vision Reasoning JSON Fastest and cheapest 5.6 model — for high volumes and latency-critical applications. Azure · EU Input 0.20 Cache Read 0.02 Cache Write 0.25 Output 1.19 from 272K input tokens from 272K Input 0.40 Cache Read 0.04 Cache Write 0.50 Output 1.77
GPT-5.4 Tools Vision Reasoning JSON Frontier model for demanding reasoning, agentic workflows and complex production workloads. Azure · EU Input 2.47 Cache Read 0.247 Cache Write — Output 14.82 from 272K input tokens from 272K Input 4.86 Cache Read 0.486 Cache Write — Output 21.80
GPT-5.4 mini Tools Vision Reasoning JSON Compact GPT-5.4 variant for high throughput at low cost — solid on standard tasks. Azure · EU Input 0.75 Cache Read 0.075 Cache Write — Output 4.45
GPT-OSS 120B Tools Reasoning JSON OpenAI's open-weight model — near-frontier reasoning with full deployment control. AKI.IO · DE Input 0.15 Cache Read — Cache Write — Output 0.55
Whisper Audio OpenAI's speech recognition — robust transcription across many languages and accents. Azure · EU 0.0065 € per minute
Get started.
Test selected models right away with €5 starter credit — every model from the first top-up. If you have questions first, talk to our sales team.
import os
from openai import OpenAI
client = OpenAI(
# one line, that's all
base_url="https://api.noirdoc.de/v1",
api_key=os.environ["NOIRDOC_API_KEY"],
)
response = client.chat.completions.create(
model="gpt-5-mini", # provider & model
messages=[{"role": "user", "content": "Hello"}],
)