Model as a Service –
Central AI model management
with German proxy
GDPR-compliant access to top models
The use of powerful AI models such as OpenAI GPT-6 Astra, Google Gemini 3.7 Flash, Anthropic Claude Fable 5.1, or xAI Grok 4.6 presents companies with new challenges: data protection, budget control, technical access, and model diversity. With partimus’ MaaS Proxy, you can access leading models via a central API—secured by German infrastructure, flexible budget management, and optional hosting for open-source solutions. Develop, scale, and integrate AI applications with maximum security and control—without waitlists, hidden costs, or compliance risks.
- GDPR-compliant hosted in Germany
- Central budget control & API access
- Ready to go immediately - no waiting time
Security, comfort and strategic
freedom in model selection
With the MaaS Proxy from partimus, you can create a central, secure and efficient basis for the productive use of modern AI models in your organization. The solution not only simplifies the connection of different applications to different language models, but also creates clarity in terms of costs, governance and data protection. Standardized interfaces, full control over the data flow and the ability to integrate your own models create a flexible and future-proof infrastructure – optimized for business use.
🏛️ German proxy service
GDPR-compliant, uniform access via German data centers.
🇩🇪 Local hosting available
Open-source models (e.g., DeepSeek V4, GLM 5.3) can be operated entirely in Germany.
🇪🇺 Choice between EU and US models
For all premium models available in the EU, we offer a choice between EU and US versions.
🎯 Central budget management with spending limits
Daily, weekly and monthly limits per team or developer. Transparent euro prices with no hidden costs.
🔧 One API for all models
OpenAI-compatible interface: Switch between GPT-6 Astra, Grok 4.6, Claude Fable 5.1, or Qwen3-Coder without changing any code.
⚡ Immediate availability
No waiting lists or additional registration – get started in minutes with prototypes up to production.
🛡️ Enterprise-grade security
ISO 27001-certified infrastructure, end-to-end encryption and central access controls.
🧾 Payment by invoice or direct debit
Pay by invoice or use the direct debit procedure. Simple and convenient.
Wide range of models - flexible and future-proof
The MaaS Proxy supports a wide range of powerful language models—from open-source LLMs such as Llama, Qwen, and DeepSeek to models from OpenAI, Google, and Anthropic. You decide which models to use, prioritize, or block—based entirely on your use case and your organization’s requirements. You can integrate your own models as well as external APIs, allowing you to flexibly customize and expand your AI infrastructure at any time. New releases can be centrally managed and deployed—without the need for manual adjustments in line-of-business applications.
- GPT-5.6 Sol (OpenAI) – The flagship model for complex tasks, designed for very long contexts
- GPT-5.6 Terra (OpenAI) – A well-rounded, general-purpose model from the GPT-5.6 series
- Claude Fable 5.1 (Anthropic) – The most powerful Claude model for complex analyses
- Claude Opus 5 (Anthropic) – A reasoning model for multi-step workflows
- Claude Sonnet 5 (Anthropic) – A balance between performance and cost, designed for continuous, productive use
- Gemini 2.5 Pro (Google) – Thinking model for analysis and content creation
- Grok 4.6 (xAI) – Real-time reasoning with up-to-date world knowledge
- Qwen3 Coder (Alibaba) – 480-parameter MoE for agent-based programming
- Kimi K2.7 Code (Moonshot AI) – Optimized for code generation, refactoring, and debugging
- GPT OSS 120B (OpenAI) – Open-source reasoning
- Kimi K3 (Moonshot AI) – Agentic intelligence for multistage tasks
- DeepSeek V4 Pro (DeepSeek) – Cost-effective reasoning model
- GLM-5.3 (Z.AI) – Hybrid reasoning
- Qwen3.5-397B-A17B (Alibaba) – 397-B MoE model with 17-B active parameters
- Muse Glimmer 30B (Meta) – Compact model for simple, high-volume tasks
Test the partimus MaaS Proxy without obligation
Benefit from centralized model management, GDPR-compliant access to leading LLMs, and maximum control over your AI usage—without any technical barriers to entry. With the partimus MaaS Proxy, you can integrate language models such as OpenAI, DeepSeek, or Llama into your systems via a unified interface—securely hosted in Germany.
Uniform API gateway for all models
The partimus MaaS Proxy is designed so that you can quickly and easily add AI functionality to your existing applications. Thanks to standardized REST API, OpenAI-compatible endpoints and full multi-provider support, the proxy can be integrated into existing tools, backends or user interfaces – without adapting the model API. Self-hosted LLMs can also be seamlessly integrated.
All requests run via a central gateway service in German data centers with the highest security and data protection standards. This gives you full control over data flow, security and infrastructure - GDPR-compliant and high-performance.
Use a wide range of AI providers such as OpenAI, Google, xAI, Alibaba and others with a single access point. The proxy takes care of routing and administration - without any adjustments to the respective API.
Deploy open-source models such as Llama, DeepSeek, or Qwen directly within your own infrastructure. The proxy enables seamless integration of local models using the same interface as that provided by cloud providers.
Keep your spending under control - with automatic budget allocation per team, model or application. Comprehensive analytics and reports help you to analyze and optimize usage transparently.
Code example
import openai
client = openai.OpenAI(
api_key="",
base_url="https://maas.partimus.com/v1"
)
response = client.chat.completions.create(
model="gpt-6", # or grok-4.6, claude-fable-5-1, qwen3-coder-480b-a35b-instruct, gemini-3.7-flash, etc.
messages=[{"role": "user", "content": "Hallo Welt!"}]
)
print(response.choices[0].message.content) Flexible pricing models for every scenario
Whether you are a startup, research team or corporation – our pricing adapts to your needs. Choose between usage-based billing per token, flat-rate monthly plans for specific providers or individually agreed enterprise models. You only pay for what you actually use – with full cost control and a clear overview. Conveniently by invoice or direct debit.
- Development Teams
- AI credit
- Proxy access
- Analytics
- Team management
- Support
- Custom SLAs
- Local hosting
- Dedicated Infrastructure
- Price
*All prices are monthly/non-recurring fees plus VAT. Our offer is aimed exclusively at business customers.
**AI Credit is credited monthly or deducted on a pro-rata basis for prepaid accounts.
Prepaid: Minimum top-up of 250 EUR. Payment by invoice or SEPA direct debit.
Unless otherwise agreed, unused credit expires 12 months after the date of the top-up. Upon expiration of this period, the right to use the corresponding credit expires.
Model prices
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| openai/gpt-6-astra | €21.00 | €78.60 |
| openai/gpt-6-astra (<272k context) | €10.50 | €52.40 |
| openai/gpt-5.6-sol | €12.76 | €57.09 |
| openai/gpt-5.6-terra | €5.17 | €20.79 |
| openai/gpt-5.6-luna | €0.55 | €2.31 |
| openai/gpt-5.6-sol (<272k context) | €6.38 | €38.06 |
| openai/gpt-5.6-terra (<272k context) | €5.17 | €20.79 |
| openai/gpt-5.6-luna (<272k context) | €0.33 | €1.54 |
| openai/gpt-5-mini | €0.33 | €2.64 |
| openai/gpt-oss-20b | €0.10 | €0.40 |
| openai/gpt-oss-120b | €0.20 | €0.70 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| moonshotai/Kimi-K3 | €3.20 | €15.80 |
| moonshotai/kimi-k2.7-code | €1.00 | €4.20 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| meta/muse-glimmer-30b | €0.40 | €1.60 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| gemini/gemini-3.7-flash | €0.80 | €4.00 |
| gemini/gemini-3.6-flash | €1.60 | €7.90 |
| gemini/gemini-3.5-flash-lite | €0.40 | €2.70 |
| gemini/gemini-2.5-pro | €2.97 | €17.38 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| xai/grok-4.6 | €2.10 | €6.30 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| qwen/qwen3-coder-480b-a35b-instruct | €0.30 | €1.10 |
| qwen/qwen3-235b-a22b-thinking | €0.20 | €0.70 |
| qwen/qwen3.5-397b-a17b | €0.60 | €3.60 |
| qwen/qwen3.6-35b-A3b | €0.30 | €1.10 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| zai/GLM-5.3 | €1.50 | €4.70 |
| zai/GLM-5.3-Flash | €0.20 | €0.60 |
| zai/glm-5.2 | €1.50 | €4.70 |
| zai/glm-5.2-instant | €1.76 | €5.28 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| deepseek/deepseek-v4-pro | €1.90 | €3.70 |
| deepseek/deepseek-v4.1-flash | €0.40 | €1.30 |
| deepseek/deepseek-v4-flash | €0.20 | €0.30 |
| Model | Cost per 1M input tokens | Cost per 1M output tokens |
|---|---|---|
| anthropic/claude-fable-5-1 | €10.50 | €52.40 |
| anthropic/claude-fable-5 | €10.50 | €52.40 |
| anthropic/claude-opus-5 | €5.30 | €26.20 |
| anthropic/claude-sonnet-5 | €3.20 | €15.80 |
| anthropic/claude-haiku-4.5 | €1.10 | €5.30 |
| anthropic/claude-sonnet-4.6 | €3.20 | €15.80 |
| anthropic/claude-opus-4.8 | €5.30 | €26.20 |
The partimus AI introductory workshop
Would you like to use modern AI tools safely, efficiently and practically in your day-to-day work? In our workshop, you will not only receive an in-depth overview of current language models and the use of central proxy access, but also get to know tried-and-tested tools such as RooCode and Chatbox. Together, we will implement individual use cases and show you how to integrate AI into your processes smoothly and in compliance with data protection regulations.
In interactive exercises, you will actively work with the tools, create your own chatbots and automated workflows and benefit from concrete best practices for cost and budget control. After just one day, you will be ready to use AI responsibly and productively in your company.
Book your workshop now and get your team ready for the use of AI.
Contents of the workshop
Basics of current language models, possible applications and trends.
Presentation of proven tools for collaborative coding and workflow automation.
Implement your own use cases and experience tools in practice.
Clearly understandable guidelines for implementation in everyday working life.
Practical tips for secure and efficient use.
Legally compliant handling of AI in your company.
Strategies for cost optimization and resource control.
Your advantages at a glance
- Practical online webinar (1 day) for developers and users
- Introduction to modern AI models and enterprise applications
- Step-by-step guide to using the AI proxy and budget management
- Training in best practices for production, security and compliance
- Active tool demos: try out using RooCode and Chatbox live
- Independent exercises & individual questions directly in the workshop
- Handout with all workflows & checklists for direct use in your team
- Certificate of participation for all participants
Maximum security & full control
The MaaS Proxy was specially developed to meet the highest data protection requirements. All requests are processed via our infrastructure in German data centers – without detours via third countries. Comprehensive logging, audit and control mechanisms allow you to keep track of data flows and access at all times. On request, sensitive models can also be hosted completely locally – for full sovereignty over your AI applications.
with German data centers and logging/monitoring.
for maximum data sovereignty.
for our information security management system.
with strict authorization and role management.
Wide range of applications for your teams
With the MaaS Proxy, you can provide powerful AI models specifically where they create real added value. Whether development, support or marketing – your entire team benefits from centrally controlled, data protection-compliant access to leading models.
Create clean code automatically, get support with debugging or generate technical documentation - directly from your development environment. Models such as Qwen3 Coder or GPT OSS 120B can be seamlessly integrated via the proxy.
Use advanced AI models such as GPT-6 Astra, Claude Fable 5.1, Gemini 3.7 Flash, or Grok 4.6 to create dynamic, multilingual chatbots. Answers to current questions are generated in real time from linked knowledge sources.
Create creative, brand-compliant content in seconds—from blog posts to social media to product descriptions. Gemini supports your editorial processes with compelling quality and consistent style.
Analyze large amounts of data quickly and efficiently using models such as DeepSeek V4 or GLM 5.3—without the need for expensive in-house GPU infrastructure. This provides you with well-founded insights for better decision-making in your day-to-day business operations.
Get started with partimus
Take the opportunity to optimize your IT infrastructure and drive your business forward – contact us today to find out more about the tailored benefits partimus can offer you.
Questions and answers about the MaaS Proxy
Are you wondering how the MaaS proxy fits into your infrastructure, which models you can integrate or what the data protection situation is? Our FAQ will give you a quick overview of the most important topics. If you have individual questions, we will be happy to answer them personally by e-mail or telephone.
What advantages does the MaaS Proxy offer compared to direct API use with OpenAI, Google & Co.?
The MaaS Proxy provides you with a central platform for managing all AI access – regardless of the provider. You benefit from standardized authentication, central monitoring, cost control, GDPR compliance and the option to also securely integrate local models.
Can we integrate our own open source models or rely exclusively on cloud providers?
Both are possible. The proxy supports both common provider models and the local operation of your own open source models – e.g. on an inference solution from partimus or on your own infrastructure.
How does the proxy help to meet data protection and compliance requirements?
The entire proxy service is operated in German data centers and is ISO 27001-certified. You retain full control over data flows, usage rights and storage locations – including logging and role-based access.
Does the use of the proxy also make sense for smaller teams or individual departments?
Yes, the flexible cost control, simple integration into existing systems and central model management in particular make the proxy suitable for smaller organizational units. This allows you to scale without having to set up a complex infrastructure.
How does billing work when using multiple models and providers?
You’ll receive a consolidated monthly statement for all requests—regardless of whether they were sent to OpenAI, Google, DeepSeek, or a local model. Budgets and limits can be flexibly set for each team, project, or user.
We will be happy to advise you!