
The DeepSeek R1-0528 release introduces a significant improvement to the reasoning capabilities of language models by making it easier to access "thinking" mode without requiring complex prompt engineering or pre-pending thinking tokens. This enhanced model provides transparent, step-by-step reasoning that's particularly valuable for educational applications, complex problem solving, and scenarios where transparency in AI decision-making is crucial.

In this post, we'll explore how to deploy the **DeepSeek-R1-0528-Qwen3-8B** model using **vLLM** on Vast.ai's cloud GPU platform, leveraging the new **qwen3 reasoning parser** that simplifies access to the model's internal thinking process.

---

## The Open Source AI Revolution

DeepSeek R1-0528 represents more than just a technical advancement—it's a significant step toward democratizing advanced AI capabilities. As an **open source model**, it makes these reasoning and advanced AI capabilities more accessible.

Unlike closed-source models that require vendor dependence, usage costs, and data privacy concerns, DeepSeek R1-0528 enables **complete control** over your AI infrastructure. Organizations can now deploy advanced reasoning capabilities locally, maintaining data sovereignty while avoiding variable per-token pricing.

This democratization means organizations can access cutting-edge AI without massive infrastructure investments. Advanced reasoning capabilities are no longer exclusive to large tech companies—anyone with GPU access can deploy their own reasoning-capable AI system.

---

## What Makes DeepSeek R1-0528 Special?

The DeepSeek R1-0528 release addresses a key limitation in previous DeepSeek models: the complexity of accessing the model's internal reasoning process. Previously, users needed to manually prepend "thinking" tokens to model outputs or use complex prompt engineering to see step-by-step reasoning.

---

## Deploying DeepSeek R1-0528 on Vast.ai

### Step 1: Set Up Vast.ai Environment

First, install the Vast CLI and configure your API key:

```bash
pip install --upgrade vastai

export VAST_API_KEY="your_vast_api_key"
vastai set api-key $VAST_API_KEY
```

### Step 2: Find Suitable Hardware

The DeepSeek-R1-0528-Qwen3-8B model requires specific hardware capabilities:

-   **Minimum 24GB GPU RAM** for model weights and KV cache
-   **Single GPU configuration** (sufficient for the 8B parameter model)
-   **Static IP address** for stable API endpoint hosting
-   **Direct port access** for the vLLM OpenAI-compatible server
-   **60GB+ disk space** for model storage and dependencies

Search for suitable instances:

```bash
vastai search offers "compute_cap >= 750 \
gpu_ram >= 24 \
num_gpus = 1 \
static_ip = true \
direct_port_count >= 1 \
verified = true \
disk_space >= 60 \
rentable = true"
```

### Step 3: Deploy the Model

Deploy using the vLLM OpenAI-compatible server with the qwen3 reasoning parser:

```bash
export INSTANCE_ID="your_instance_id"

vastai create instance $INSTANCE_ID \
--image vllm/vllm-openai:latest \
--env '-p 8000:8000' \
--disk 60 \
--args --model deepseek-ai/DeepSeek-R1-0528-Qwen3-8B \
--max-model-len 4096 \
--reasoning-parser qwen3
```

**Key deployment parameters:**

-   `--reasoning-parser qwen3`: Enables the simplified reasoning parser
-   `--max-model-len 4096`: Accommodates longer reasoning sequences
-   `--tensor-parallel-size 1`: Single GPU configuration
-   `--gpu-memory-utilization 0.90`: Optimized memory usage

### Step 4: Get Instance Details and Test Connection

Navigate to the [Instances tab](https://cloud.vast.ai/instances/) in the Vast AI Console and locate your instance. Click the IP address button to view the forwarded ports.

Test the connection with a simple curl command:

```bash
export VAST_IP_ADDRESS="your_ip_address"
export VAST_PORT="your_port"

curl -X POST http://$VAST_IP_ADDRESS:$VAST_PORT/v1/completions \
-H "Content-Type: application/json" \
-d '{"model": "deepseek-ai/DeepSeek-R1-0528-Qwen3-8B", "prompt": "Hello, how are you?", "max_tokens": 50}'
```

---

## Testing the Model's Reasoning Capabilities

### Setup Python Client

```python
import time
from openai import OpenAI

# Replace with your actual Vast.ai instance details
VAST_IP_ADDRESS = "your_ip_address"
VAST_PORT = "your_port"

client = OpenAI(
    api_key="DUMMY",
    base_url=f"http://{VAST_IP_ADDRESS}:{VAST_PORT}/v1"
)
```

### Basic Functionality Test

```python
start = time.time()
test_resp = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
    messages=[{"role": "user", "content": "Hello"}]
)
elapsed = time.time() - start

print("Response:", test_resp.choices[0].message.content)
print(f"Elapsed time: {elapsed:.4f} s")
```

### Reasoning Capability Test

This is where DeepSeek R1-0528 truly shines. The model naturally provides step-by-step reasoning without complex prompt engineering:

```python
reasoning_prompt = """
Write the History of Paris in a paragraph, show your thinking
step by step first before paragraph and then write the history
in a paragraph
"""

start = time.time()
resp = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
    messages=[{"role": "user", "content": reasoning_prompt}]
)
elapsed = time.time() - start

print("Model output:\n", resp.choices[0].message.content)
print(f"Elapsed time: {elapsed:.3f} s")
```

**Expected output structure:**

1. **Thinking process**: The model shows its step-by-step reasoning
2. **Final answer**: A well-structured paragraph about Paris's history

### Mathematical Reasoning Test

```python
math_prompt = "If 8 × 7 = ?, show your work step by step."

resp = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
    messages=[{"role": "user", "content": math_prompt}]
)

print("Mathematical reasoning:\n", resp.choices[0].message.content)
```

The model will naturally break down the problem, show intermediate steps, and arrive at the solution with clear reasoning traces.

---

## Key Benefits of DeepSeek R1-0528

### 1. **Simplified Reasoning Access**

No need to prepend thinking tokens or use complex prompt engineering. The `qwen3` reasoning parser automatically handles the extraction and presentation of reasoning steps.

### 2. **Educational Value**

Perfect for educational applications where students need to understand the problem-solving process, not just the final answer.

### 3. **Transparency and Trust**

Users can see exactly how the model arrives at its conclusions, building trust in AI-powered decision-making systems.

### 4. **Cost-Effective Deployment**

Running on Vast.ai provides significant cost savings compared to major cloud providers while maintaining high performance.

### 5. **Easy Integration**

OpenAI-compatible API ensures seamless integration with existing applications and workflows.

---

## Use Cases

### Educational Applications

-   Step-by-step problem solving for mathematics, science, and logic
-   Teaching critical thinking and reasoning skills
-   Creating interactive learning experiences

### Business Intelligence

-   Transparent decision-making in automated systems
-   Audit trails for AI-powered recommendations
-   Complex analysis with clear reasoning paths

### Research and Development

-   Understanding model behavior and capabilities
-   Testing reasoning patterns across different domains
-   Developing more interpretable AI systems

---

## Conclusion

The DeepSeek R1-0528 release represents a significant step forward in making AI reasoning more accessible and transparent. By combining this powerful model with Vast.ai's cost-effective GPU infrastructure and the simplified qwen3 reasoning parser, developers can now easily deploy transparent, reasoning-capable AI systems.

Whether you're building educational tools, business intelligence systems, or research applications, DeepSeek R1-0528 on Vast.ai provides an excellent foundation for applications that require clear, step-by-step reasoning capabilities.

The elimination of complex prompt engineering for accessing reasoning modes makes this model particularly valuable for production applications where transparency and interpretability are key requirements.

---

## Useful Resources

-   [DeepSeek R1-0528 Model on Hugging Face](https://huggingface.co/deepseek-ai/DeepSeek-R1-0528-Qwen3-8B)
-   [vLLM Documentation](https://docs.vllm.ai/)
-   [Vast.ai GPU Marketplace](https://vast.ai)
-   [OpenAI Python SDK](https://github.com/openai/openai-python)

---

_Ready to deploy transparent, reasoning-capable AI? Get started with DeepSeek R1-0528 on Vast.ai today! 🚀_


Deploy DeepSeek R1-0528 on Vast.ai to access transparent, step-by-step AI reasoning without prompt tricks—ideal for education, research, and business use.

DeepSeek R1-0528: Enhanced Reasoning with Simplified Thinking Mode on Vast.ai


## Introduction

With the release of o1 family of models, reasoning has taken the AI world by storm. Instead of immediately answering a prompt, reasoning is where a model is trained to first output it's "thinking" for its answer. This increases performance on the most difficult tasks today, and since it is trained into the model, it is better than asking it to record its "chain of thought" via a prompt.

R1 takes this even further. We now have Open Source models for this type of task, which means:

1. We can retain the "thinking" token outputs which we couldn't do beforehand. This lets us create datasets for finetuning.
2. We can easily fine-tune these models on our own tasks
3. We can run them on more affordable compute to drastically reduce the cost!

This implementation combines three key components: A distilled DeepSeek model for reasoning transparency, a Vast Template for vLLM's optimized inference server for efficient deployment, and Langchain for parsing reasoning tokens versus normal output tokens. Together, they create a production-ready system that can handle both the technical demands of serving a large language model and the practical needs of processing its unique output format.

Vast.ai provides an ideal platform for this setup, offering the necessary GPU resources at a fraction of traditional cloud costs. Its marketplace model and simple Docker integration make it particularly well-suited for deploying and scaling language models, allowing developers to focus on building applications rather than managing infrastructure. The Templates feature from Vast allows for a repeatable deployment with little additional configuration.

This guide demonstrates how to deploy [deepseek/DeepSeek-R1-Distill-Qwen-32B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-32B) with Vast's templates and integrate them with Langchain for advanced processing capabilities. We'll show you how to leverage vLLM's optimized inference server and create custom parsers to handle DeepSeek's distinctive output format. Feel free to follow along in this [companion notebook](https://nbviewer.org/urls/bitbucket.org/!api/2.0/snippets/jsbcannell/xq8zG5/2cf7d8758235fa5176d9aee8b8d3885128c0109c/files/serving_deepseek_on_vast.ipynb)

## Setting Up the Environment

Before we can deploy our model, we need to set up our Vast.ai environment. First, install the Vast SDK:

```python
%%bash
pip install --upgrade vastai
```

Set up your Vast API key (available from your [Account Page](https://cloud.vast.ai/account/)):

```python
%%bash
# Here we will set our api key
export VAST_API_KEY="<your-key-here>"
vastai set api-key $VAST_API_KEY
```

### Choosing the Right Hardware

Now we are going to search for GPUs on Vast.ai to run the DeepSeek-R1-Distill-Qwen-32B model. This model requires specific hardware capabilities to run efficiently with vLLM's optimizations. Here are our requirements:

1. A minimum of 80GB GPU RAM to accommodate:

    - DeepSeek model weights (32B Parameters)
    - KV Cache for handling of extra long output token lengths

2. A single GPU configuration, as DeepSeek-R1-Distill-Qwen-32B can be efficiently served on one GPU:
   Note: Multi-GPU configurations are supported if higher throughput is needed.

3. A static IP address for:

    - Stable API endpoint hosting
    - Consistent client connections
    - Reliable Langchain integration

4. At least one direct port that we can forward for:

    - vLLM's OpenAI-compatible API server
    - External access to the model endpoint
    - Secure request routing

5. At least 120GB of disk space to hold the model and other things we might like to download

Here's how to search for suitable instances on Vast.ai:

```python
%%bash
vastai search offers "compute_cap >= 750 \
gpu_ram >= 80 \
num_gpus = 1 \
static_ip = true \
direct_port_count >= 1 \
verified = true \
disk_space >= 120 \
rentable = true"
```

### Deploying the Server via Vast Template

We'll use vLLM's OpenAI-compatible server to deploy the DeepSeek model. This setup provides an OpenAI-compatible API endpoint that works seamlessly with existing tools and libraries.

We will do this with a template that:

1. Uses `vllm/vllm-openai:latest` docker image. This gives us an OpenAI-compatible server.
2. Forwards port `8000` to the outside of the container, which is the default OpenAI server port
3. Forwards `--model deepseek-ai/DeepSeek-R1-Distill-Qwen-32B --max-model-len 8192  --enforce-eager` on to the default entrypoint (the server itself)
4. Uses `--tensor-parallel-size 1` by default.
5. Uses `--gpu-memory-utilization 0.90` by default
6. Ensures that we have 120 GB of Disk space

```python
%%bash
export INSTANCE_ID= #insert instance ID
vastai create instance $INSTANCE_ID --disk 120 --template_hash eda062b3e0c9c36f09d9d9a294405ded
```

### Verify Setup

After deployment, verify that your server is running correctly with a simple curl test:

```python
%%bash
export VAST_IP_ADDRESS="<your-ip-here>"
export VAST_PORT="<your-port-here>"
curl -X POST http://$VAST_IP_ADDRESS:$VAST_PORT/v1/completions \
     -H "Content-Type: application/json" \
     -d '{
           "model": "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
           "prompt": "Hello, how are you?",
           "max_tokens": 50
         }'
```

Note: You can find your instance's IP address and port in the [Instances tab](https://cloud.vast.ai/instances/) of the Vast AI Console.

## Implementing Custom Output Parsing

### Install Required Dependencies

```python
%%bash
pip install --upgrade langchain langchain-openai openai
```

### Creating a DeepSeek Output Parser

DeepSeek models output responses in a unique format with separate thinking and response sections. The thinking section, wrapped in `<think>` tags, contains the model's reasoning process, while everything after the closing tag represents the final response. This separation is valuable for understanding how the model reaches its conclusions.

Let's create a custom parser that can handle this format:

```python
from typing import Optional, Tuple
from langchain.schema import BaseOutputParser

class R1OutputParser(BaseOutputParser[Tuple[Optional[str], str]]):
    """Parser for DeepSeek R1 model output that includes thinking and response sections."""

    def parse(self, text: str) -> Tuple[Optional[str], str]:
        """Parse the model output into thinking and response sections.

        Args:
            text: Raw text output from the model

        Returns:
            Tuple containing (thinking_text, response_text)
            - thinking_text will be None if no thinking section is found
        """
        if "</think>" in text:
            # Split on </think> tag
            parts = text.split("</think>")
            # Extract thinking text (remove <think> tag)
            thinking_text = parts[0].replace("<think>", "").strip()
            # Get response text
            response_text = parts[1].strip()
            return thinking_text, response_text

        # If no thinking tags found, return None for thinking and full text as response
        return None, text.strip()

    @property
    def _type(self) -> str:
        """Return type key for serialization."""
        return "r1_output_parser"
```

Our parser does several important things:

1. It looks for the `</think>` tag to identify the boundary between thinking and response
2. It extracts and cleans up the thinking section by removing the `<think>` tags
3. It handles cases where no thinking section is present by returning None
4. It ensures both sections are properly stripped of extra whitespace

### Setting Up the Model Chain

Now that we have our parser, we need to set up a processing chain that connects our model to the parser. This chain will:

-   Send requests to our deployed model
-   Process the responses through our custom parser
-   Return the separated thinking and response sections

Here's how we set it up:

```python
from langchain_openai import ChatOpenAI
from langchain.prompts import ChatPromptTemplate
from langchain.schema.runnable import RunnablePassthrough

# Initialize the model with your Vast instance details
VAST_IP_ADDRESS="<your-ip-here>"
VAST_PORT="<your-port-here>"

openai_api_key = "EMPTY"
openai_api_base = f"http://{VAST_IP_ADDRESS}:{VAST_PORT}/v1"

model = ChatOpenAI(
    base_url=openai_api_base,
    api_key=openai_api_key,
    model_name="deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
    max_tokens=8000,
    temperature=0.7
)

# Create prompt template
prompt = ChatPromptTemplate.from_messages([
    ("user", "{input}")
])

# Create parser
parser = R1OutputParser()

# Create chain
chain = (
    {"input": RunnablePassthrough()}
    | prompt
    | model
    | parser
)
```

Let's break down what each component does:

-   The `ChatOpenAI` instance connects to our deployed model using the OpenAI-compatible API
-   The `ChatPromptTemplate` formats our input messages
-   The `RunnablePassthrough` ensures our inputs flow through the chain correctly
-   Finally, our custom parser processes the model's output

### Example Usage

Let's test our setup with a prompt that can demonstrate the model's reasoning capabilities:

```python
prompt_text = "Explain quantum computing to a curious 10-year-old who loves video games."

thinking, response = chain.invoke(prompt_text)
print("\nTHINKING:\n")
print(thinking)
print("\nRESPONSE:\n")
print(response)
```

#### Output:

```
THINKING:

Okay, so I need to explain quantum computing to a 10-year-old who loves video games. Hmm, let's break this down. First, I should think about what a 10-year-old knows. They understand basic concepts like computers, games, maybe some video game terms like pixels or bits. They also enjoy stories or analogies that relate to their interests, so using video games as a context makes sense.

Quantum computing is a complex topic, so I need to simplify it. I remember that classical computers use bits, which are 0s and 1s. Quantum computers use qubits, which can be both 0 and 1 at the same time. That's a key difference. Maybe I can compare it to something in a video game. Maybe like a character that can be in two places at once, or a power-up that gives multiple abilities simultaneously.

I should also mention superposition and entanglement. Superposition is when a qubit can be in multiple states at once, and entanglement is when qubits are connected, so the state of one affects the others. Maybe relate that to something like a power-up that affects multiple things at once or a team of characters working together.

Another important point is that quantum computers can solve certain problems much faster than classical computers. Maybe compare it to a game where you can unlock a level or defeat a boss much quicker with a special power-up. But I should also note that not all problems are better solved by quantum computers, just specific ones.

Putting it all together, the explanation should start with what a classical computer does, then introduce qubits as special bits, explain their unique properties using video game analogies, and then talk about the advantages and limitations. I should keep the language simple, use terms they're familiar with, and make it engaging.

Let me think of an example. Maybe a game where you have to find a key in a maze. A classical computer would check each path one by one, while a quantum computer could check all paths at the same time, making it faster. Or maybe a power-up that lets you see all possible paths at once.

I also want to make it clear that it's not just faster, but it's a different way of computing. Maybe use the idea of magic or special abilities in a game that allow for unique capabilities.

I should avoid technical jargon and keep sentences short. Maybe use a narrative style, like telling a story about a game where quantum computing helps the player win. That could make it more relatable and fun.

So, to sum up, I'll start by comparing classical computers to video game rules, then introduce qubits as special bits that can do multiple things at once, explain superposition and entanglement with game-related examples, and discuss the speed advantages and specific applications. I'll wrap it up by connecting it back to video games, showing how quantum computing could make games more complex or fun in the future.

I think that's a solid approach. Now, I'll structure this into a clear, concise explanation using these points and analogies.

RESPONSE:

Quantum computing is like having a supercharged video game character! Imagine you're playing a game where you need to find a hidden treasure in a maze. A regular computer (like a normal game character) would check each path one by one, which could take a long time. But a quantum computer is like a special character that can check all paths at the same time, making it super fast to find the treasure!

In video games, you sometimes get power-ups that give you special abilities. Quantum computers use something called qubits, which are like those power-ups. Unlike regular bits in a computer that can only be 0 or 1, qubits can be both at the same time. This is called superposition. It's like having a character that can be in two places at once!

Also, qubits can work together in a magical way called entanglement. If one qubit changes, the others change too, no matter how far apart they are. It's like having a team of characters where each member's actions affect the others instantly.

Quantum computers are especially good at solving tricky puzzles or problems that would take a regular computer forever. But they're not better at everything—just certain tasks. So, in the game, your special character would shine in specific levels, making them super fun and challenging!

One day, quantum computing might help create even more exciting games, with puzzles that seem impossible to solve without those special powers. So, think of quantum computing as the ultimate power-up for solving complex problems in the blink of an eye!
```

This example shows how the model:

1. First works through its reasoning in the thinking section
2. Then provides a clear, final response
3. Keeps these two aspects separate for better analysis and control

## Key Features

-   **Separated Thinking and Response**: DeepSeek's unique output format provides insight into the model's reasoning process
-   **Custom Output Parsing**: Langchain integration enables structured parsing of the model's unique output format

## Conclusion

This guide has demonstrated how to deploy and serve the DeepSeek-R1-Distill-Qwen-32B model on Vast.ai. We've shown how to:

-   Set up a cost-effective GPU instance with the right specifications
-   Easilly deploy the model using Vast's Templates with vLLM
-   Create a custom Langchain parser to handle DeepSeek's unique thinking/response format
-   Integrate everything through an OpenAI-compatible API

With this setup on Vast.ai's GPU marketplace, you can now build powerful AI applications with DeepSeek while minimizing both infrastructure complexity and compute costs. The platform's combination of flexible GPU options and simple Docker-based deployment makes it an ideal foundation for serving and scaling language models without the expensive cost model of traditional cloud providers.


Deploy DeepSeek-R1-Distill-Qwen-32B on Vast.ai for cost-effective AI reasoning. Use vLLM, Langchain, and custom parsing to extract structured model insights efficiently.

Serving DeepSeek Models on Vast.ai with vLLM and Langchain!


# January 2025 Product Update

We're back with our first product update of 2025!

This month, we're pleased to announce we've added support for [DeepSeek R1](https://vast.ai/article/deepseek-r1-open-source-disruptor-or-overhyped-upstart), so you can easily run this powerful new LLM through our platform. Our DeepSeek R1 / Ollama WebUI Guide and Ollama template streamline the process, giving you a seamless way to leverage DeepSeek's capabilities for your own projects.

As your trusted partner in cloud GPU resources, we're constantly working to enhance every aspect of our platform. As with every update, we prioritize addressing any technical hiccups that may affect your experience on Vast. Here are some of the latest bugs we've fixed since our last update:

- Teams account switching bugs.
- Teams invite bugs and UX issues.
- Fixed earning page issues related to earning display and Stripe.
- Erroneous client build errors showing on machines.
- Bandwidth speeds misreporting.
- Instances open button issues.

## NVIDIA Cloud GPU Updates

Here at Vast.ai, we're always looking for ways to make your experience even better, whether you're exploring cutting-edge AI models or running established workflows.

### Highlights of the latest enhancements:

- Added [](https://docs.vast.ai/ollama-webui)[DeepSeek R1 / Ollama + WebUI Guide](https://docs.vast.ai/ollama-webui) and [](https://cloud.vast.ai/template/readme/5d2c8df8e0d0ab7d1a7a3ed0f9105637)[Ollama Template](https://cloud.vast.ai/template/readme/5d2c8df8e0d0ab7d1a7a3ed0f9105637).
- Updated [](https://cloud.vast.ai/account/)[Earning Page](https://cloud.vast.ai/account/) UI, including visualizations of referral program earnings, template performance, and host earnings.
- [Docs Overhaul](https://docs.vast.ai/) – We cleaned up API Docs, replatformed and fixed Docs pages UI/UX and search issues, and added new guides.
- [Vast.ai Base Docker Image](https://github.com/vast-ai/base-image) – This clonable repo allows template creators to quickly and easily build on top of a generic Vast.ai Docker image.
- [Audit Logs](https://cloud.vast.ai/account/) added to web UI – Users can filter and view history of actions taken on their account.
- (Rolling out) Onboarding Flow – Helpful sidebar checklist that guides new users (including referrals) through all the steps necessary to launch their first instance.

For the full list of all our January 2025 updates, you can find the complete change log at the bottom of this page. We're excited for you to explore all the latest improvements!

## Our Commitment

Whether you're looking to get started with DeepSeek R1 or interested in exploring GPU or VM rental options, we're here to help you make the most of our platform. Got questions? We'd love to hear from you!

Send us an email anytime at [support@vast.ai](mailto:support@vast.ai), or join our Discord community for instant access to resources and helpful discussions. Let us show you how Vast.ai can power your next project!

## Change Log

### Features:

- Updated [](https://cloud.vast.ai/account/)[Earning Page](https://cloud.vast.ai/account/) UI, including visualizations of referral program earnings, template performance, and host earnings.
- [Docs Overhaul](https://docs.vast.ai/) – We cleaned up API Docs, replatformed and fixed Docs pages UI/UX and search issues, and added new guides.
- [Vast.ai Base Docker Image](https://github.com/vast-ai/base-image) – This clonable repo allows template creators to quickly and easily build on top of a generic Vast.ai Docker image.
- [Audit Logs](https://cloud.vast.ai/account/) added to web UI – Users can filter and view history of actions taken on their account.
- [Account Page](https://cloud.vast.ai/account/) UI Refresh.
- (Rolling out) Onboarding Flow – Helpful sidebar checklist that guides new users (including referrals) through all the steps necessary to launch their first instance.

### Templates & Guides:

- Added [](https://docs.vast.ai/ollama-webui)[DeepSeek R1 / Ollama + WebUI Guide](https://docs.vast.ai/ollama-webui) and [](https://cloud.vast.ai/template/readme/5d2c8df8e0d0ab7d1a7a3ed0f9105637)[Ollama Template](https://cloud.vast.ai/template/readme/5d2c8df8e0d0ab7d1a7a3ed0f9105637).
- Added [](https://docs.vast.ai/image-generation)[Image Generation Guide](https://docs.vast.ai/image-generation).
- Added [](https://docs.vast.ai/linux-virtual-machines)[Linux Virtual Machine Guide](https://docs.vast.ai/linux-virtual-machines).
- Added [](https://docs.vast.ai/linux-virtual-desktop)[Linux Virtual Desktop Guide](https://docs.vast.ai/linux-virtual-desktop).
- Added [](https://docs.vast.ai/pytorch)[PyTorch Guide](https://docs.vast.ai/pytorch).
- Added [](https://cloud.vast.ai/?ref_id=62897&template_id=0616da77103ae90486bf60b8ddcd4d9c)[NVIDIA CUDA Base Template](https://cloud.vast.ai/?ref_id=62897&template_id=0616da77103ae90486bf60b8ddcd4d9c).
- Updated A1111 Template.
- Updated PyTorch Templates.
- Updated KVM Desktop Template.

### Bug Fixes:

- Teams account switching bugs.
- Teams invite bugs and UX issues.
- Fixed earning page issues related to earning display and Stripe.
- Disk space selection mismatches and errors.
- Erroneous client build errors showing on machines.
- Bandwidth speeds misreporting.
- Machine API key saving issue.
- Various template editor UI issues related to version tags, environment vars, and ports.
- Various template listing UI issues related to loading / “load more” and card display.
- Instances open button issues.
- Security issues.


Vast.ai's January 2025 update adds support for DeepSeek R1, new guides, UI enhancements, bug fixes, and NVIDIA GPU improvements to enhance cloud experience and performance.

January 2025 Product Update


# DeepSeek R1: Open-Source Disruptor or Overhyped Upstart?

DeepSeek, a Chinese AI firm, has drawn attention in recent weeks for its release of DeepSeek R1 – a large language model (LLM) that aims to stand shoulder to shoulder with more established players. Its rapid rise has shaken up the AI race, with comparisons to OpenAI's o1 in particular.

## Try it For Yourself on Vast

Our [Ollama + WebUI](https://docs.vast.ai/ollama-webui) guide uses DeepSeek r1:70b as an example. Follow it to try out DeepSeek on our GPU cloud.

## The Disruptive Growth of DeepSeek

DeepSeek's impact isn't limited to industry insiders. It's already gaining massive popularity among everyday users. The DeepSeek mobile app, which provides a chatbot interface for DeepSeek R1, shot to the top of the Apple App Store charts within days of its release on January 20, 2025.

But the biggest reason DeepSeek is [making waves](https://techcrunch.com/2025/01/28/deepseek-everything-you-need-to-know-about-the-ai-chatbot-app/)? **The company has seemingly disrupted the prevailing wisdom that developing cutting-edge LLMs costs a ton of money.** The DeepSeek R1 LLM was developed at a fraction of the cost of comparable models from other companies – reportedly less than $6 million as opposed to hundreds of millions or even billions of dollars.

Another key factor setting it apart: **DeepSeek R1 is free and open source.** Researchers and developers can freely experiment with, modify, and deploy the model without the restrictions of proprietary alternatives.

The meteoric rise of the R1 LLM threw the U.S. financial markets into turmoil, prompting investors to rethink the valuation of companies like NVIDIA and reevaluate the AI prospects of tech giants like Microsoft and Meta. A significant stock market sell-off occurred on January 27, 2025.

Beyond its financial impact, though, what exactly is DeepSeek and how did it develop a model that's turning heads? Let's back up and take a closer look at this new player on the AI scene.

## What Is DeepSeek?

Based in Hangzhou, China, **DeepSeek is an AI development firm that focuses on creating open-source LLMs.** While it operates as an independent lab, DeepSeek is backed by High-Flyer Capital Management, a China-based quantitative hedge fund specializing in AI-driven financial strategies.

From the start, DeepSeek built its own data center clusters to train its generative AI models. U.S. export restrictions on advanced chips had an impact on operations, however. DeepSeek was forced to train its more recent models using NVIDIA H800 chips, a less powerful alternative to the H100 chips available to U.S. companies.

DeepSeek's first model was released in November 2023, and since then the firm has developed several different variations. But DeepSeek only catapulted to global prominence in January 2025, when its R1 model was released – along with a web interface, mobile application, and API access.

## DeepSeek R1: A Challenge to the Status Quo?

The DeepSeek R1 model focuses on logical inference, mathematical reasoning, and real-time problem-solving. Compared to other LLMs, **it uses a more efficient and sparse Mixture-of-Experts (MoE) architecture, where different parts of the model specialize in handling different tasks.** For every input, only a few "expert" neural networks are selected to process each piece of data.

What's really sparked buzz, however, is DeepSeek R1's ability to match performance with OpenAI's latest ChatGPT models across various tasks. So R1 is a more efficient model, developed much more cheaply, with impressive performance. It sounds almost too good to be true...

How was it trained?

According to DeepSeek, the R1 model leveraged the following training innovations:

-   **Reinforcement Learning** – Using a large-scale reinforcement learning approach without supervised fine-tuning, DeepSeek employed techniques like group relative policy optimization (GRPO) to enhance reasoning capabilities.
-   **Reward Engineering** – Instead of relying on traditional neural reward models, DeepSeek created a rule-based reward system that delivers superior results (i.e., by structuring incentives to guide the model's learning during training).
-   **Emergent Behavior Network** – DeepSeek discovered that advanced reasoning patterns can emerge on their own through reinforcement learning, without being explicitly programmed.
-   **Distillation** – DeepSeek used efficient knowledge transfer techniques and distilled knowledge from larger, more complex AI models, allowing R1 to achieve strong performance with reduced computational demands.

Not all of this is new. DeepSeek does present some algorithmic and architectural novelties – such as the sparse MoE design and the use of reinforcement learning to refine synthetic data selection and optimize reasoning pathways during training – but the company isn't necessarily the first to implement these techniques.

Still, it has combined them in a way that makes its R1 model notably efficient and competitive.

## Too Good to Be True?

Unfortunately, there is a catch. **DeepSeek R1 reportedly [relied](https://www.theverge.com/news/601195/openai-evidence-deepseek-distillation-ai-data) on synthetic data from OpenAI's GPT-4o for its training**, pushing some of the computational burden onto the latter, much less efficient model.

So while R1 is touted as being highly efficient, those claims only apply to the model itself, not to the entire process used to train it. The R1 model's lean MoE architecture means it is indeed efficient when running inference, but its training pipeline requires external synthetic data generated by other models like GPT-4o – and that external data generation is computationally expensive.

This means that a significant part of the total AI development cost and compute burden is still being offloaded onto other, more complex models. **It's a cost-saving move for DeepSeek, but it doesn't negate the need for massive specialized compute in the industry overall.**

Time will tell how sustainable DeepSeek's approach is – and what consequences it may have for the company as scrutiny over training methods increases and competition in the AI space continues to grow.

...There may even be national security concerns, as the U.S. government [eyes](https://www.cbsnews.com/news/deepseek-ai-raises-national-security-concerns-trump/) the potential for harmful foreign influence. But that's a topic for another day.

## The Bottom Line on DeepSeek R1

Regardless of how one feels about it, DeepSeek R1 has certainly made a bold entrance into the world. While its efficiency claims come with important caveats, the model remains a fascinating case study in how AI development is evolving.

Plus, its open-source license allows for commercial use, which is a pretty major draw for developers around the world.

## Want to Learn More

DeepSeek R1 is a **massive** model that will not fit on a single 8xH200 instance unless you run a quantized version. The 4bit quantization will run on a 8xH100 or 4xH200 node. The sweet spot might be the 8bit quantized model. If you are interested in trying it out, [reach out to us](https://vast.ai/contact-sales) so we can talk to you about your use case and help get it setup.

As always, we'd be happy to assist you in any way we can! We're here to support your AI experiments, providing the compute power you need to explore, test, and innovate – every day.


DeepSeek R1: Open-source LLM shaking up the AI race. Developed for under $6M, it rivals ChatGPT's performance. Try it on Vast's GPU cloud.

Posts about: Deepseek

DeepSeek R1-0528: Enhanced Reasoning with Simplified Thinking Mode on Vast.ai

Serving DeepSeek Models on Vast.ai with vLLM and Langchain!

January 2025 Product Update

DeepSeek R1: Open-Source Disruptor or Overhyped Upstart?