DeepSeek-R1:70B¶
In this article
Information
DeepSeek-R1:70B is an advanced language model with 70 billion parameters, designed for high-performance tasks and local deployment through the Ollama framework. It combines exceptional expressive power, support for complex inferences, and easy integration via Open Web UI. To function effectively, the model requires powerful computational resources—specifically modern GPUs from NVIDIA (with FP16/INT4 support) or compatible accelerators. Deployment is recommended on Ubuntu 22.04 or later versions of the OS to ensure stable operation with large models.
Main Features of DeepSeek-R1:70B¶
- High Performance: Thanks to its massive architecture with 70 billion parameters, DeepSeek-R1:70B demonstrates outstanding results in natural language generation and comprehension, ensuring accuracy and depth of responses even in complex scenarios;
- Multilingual Support: The model is trained on extensive multilingual corpora and can confidently work with dozens of languages, including Russian, English, Chinese, Spanish, French, and many others;
- Advanced Inference Modes: Supports zero-shot, few-shot, and chain-of-thought reasoning, allowing it to solve complex logical, analytical, and creative tasks without the need for further training;
- Versatile Application: DeepSeek-R1:70B efficiently handles a wide range of tasks — from generating artistic and technical text to writing and debugging code, solving mathematical problems, and analyzing structured data;
- Deep Integration: The model can easily connect to external systems via REST API or be embedded in chatbots, analytical platforms, IDEs, and corporate applications;
- Fine-tuning and Adaptation Capabilities: Supports fine-tuning and LoRA adaptation for specialized domains — such as medicine, finance, law, engineering, and scientific research;
- Ethical and Reliable: Includes built-in mechanisms for filtering toxic, harmful, or biased content, aligning with modern standards of responsible AI;
- Optimized for Local Use: Despite its scale, DeepSeek-R1:70B supports quantization (e.g., down to 4-bit), allowing it to run on servers with limited GPU memory without critical quality loss;
-
Handling Heterogeneous Data: The model efficiently processes not only standard text but also programming code, tables, JSON, XML, and other formats, making it a valuable tool in data science and automation;
-
Examples of Use:
- Intelligent Customer Support: Generating personalized, context-dependent responses in real-time;
- Education and Science: Assisting with solving complex problems, explaining concepts, generating educational materials;
- Content and Marketing: Creating creative texts, analyzing tone, generating ideas for campaigns;
- Software Engineering: Code autocompletion, refactoring, documentation, generation of unit tests.
Deployment Features¶
| ID | Compatible OS | VM | BM | VGPU | GPU | Min CPU (Cores) | Min RAM (Gb) | Min HDD/SDD (Gb) | Active |
|---|---|---|---|---|---|---|---|---|---|
| 410 | Ubuntu 22.04 GPU | - | - | + | + | 8 | 128 | 240 | ORDER |
- Installation time is 30-40 minutes along with the OS;
- The Ollama server loads and runs LLM in memory;
- Open WebUI is deployed as a web application connected to the Ollama server;
- Users interact with the LLM through the Open WebUI web interface, sending requests and receiving responses;
- All computations and data processing occur locally on the server. Administrators can configure the LLM for specific tasks through OpenWebUI tools.
System Requirements and Technical Specifications¶
- Operating System: Ubuntu 22.04;
- RAM: minimum 128 GB RAM;
- Graphics Accelerator: 2x5090 with 32 GB video memory (64 Gb summary) or other configurations. It is recommended to use A100/H100/RTX 6000 PRO
- Disk Space: Sufficient for installing the system, drivers, and the model;
- Drivers: NVIDIA drivers and CUDA for proper GPU operation;
- Video Memory Consumption: 48 GB at a 2K token context;
- Automatic Restart: Automatic container restart is set up in case of failures;
- GPU Support: Full integration with NVIDIA CUDA for maximum performance.
Getting Started After Deploying DeepSeek-R1:70B¶
After payment, an email will be sent to the address specified during registration notifying you that the server is ready. It will include the VPS IP address, as well as login and password for accessing the server and a link to access the OpenWebUI control panel. Clients of our company manage equipment through the server management panel and API — Invapi.
-
Login Data for Accessing the Server's Operating System (e.g., via SSH) will be sent to you in the email.
-
Link for Accessing Ollama Control Panel with Open WebUI Web Interface: In the webpanel tag under the Info >> Tags tab of the Invapi control panel. The exact link in the format
https://deepseek<Server_ID_from_Invapi>.hostkey.inwill be sent via email when the server is delivered.
After clicking the link from the tag webpanel, a Get started with Open WebUI login window will open, where you need to create an admin account name, email, and password for your chatbot, then press the Create Admin Account button:
Attention
After registering the first user, the system automatically assigns them an administrator role. To ensure security and control over the registration process, all subsequent registration requests must be approved in OpenWebUI from the administrator account.
Note
Detailed information on features of working with the Ollama control panel with Open WebUI can be found in the article AI Chatbot on Your Own Server.
Note
For optimal performance, it is recommended to use a GPU with more than the minimum requirement of 48 GB video memory. This ensures headroom for processing larger contexts and parallel requests. Detailed information on main Ollama settings and Open WebUI can be found in the developers' documentation of Ollama and in the developers' documentation of Open WebUI.
Ordering a Server with DeepSeek-R1:70B via API¶
To install this software using the API, follow these instructions.
Some of the content on this page was created or translated using AI.