At Novita AI, we’re excited to announce that our DeepSeek model has been upgraded to support a content length of 160k, a significant improvement over the previous 128k limit. This upgrade means you can now process and analyze even larger datasets in a single go, saving time and improving efficiency. For a limited time, new users can claim **$10 in free credits**to explore the upgraded DeepSeek models and various other LLM APIs on Novita AI.
DeepSeek-R1-0528
Introduction
DeepSeek R1 0528 was launched on May 28, 2025, as an open-source large model with approximately 685 billion parameters. It uses a Mixture-of-Experts (MoE) architecture, activating about 37 billion parameters per token during inference. The model supports a maximum context length of 128K tokens.
The model excels in chat, reasoning, coding, mathematics, and function calling, with added support for JSON output and function call interfaces, significantly enhancing its ability to handle complex tasks. It was trained on over 10 trillion tokens, including web content, code, mathematical data, and documents, with a strong focus on bilingual capabilities in English and Chinese.
Training involved traditional Reinforcement Learning from Human Feedback (RLHF) and fine-tuning methods, combined with substantial compute resources and algorithmic optimizations in the later stages. This approach prioritizes accuracy and reliability over efficiency, making the model well-suited for enterprise applications, especially those requiring complex reasoning and high precision.
Benchmark
DeepSeek-V3-0324
Introduction
DeepSeek V3 0324 was launched on March 25, 2025, as an open-source large model with approximately 671 billion parameters, activating about 37 billion parameters per token during inference. It utilizes a Mixture-of-Experts (MoE) architecture and supports a maximum context length of 160K tokens, making it highly capable of handling ultra-long inputs and generating extended outputs.
The model is designed to excel in tasks such as multi-turn conversations, in-depth reasoning, coding generation, and mathematical problem-solving. Enhanced multilingual support, particularly for Chinese, and its text-to-text multimodal capability further extends its versatility. The training data includes 14.8 trillion tokens of diverse sources, such as web content, codebases, and technical documents, ensuring a robust knowledge base for complex applications.
DeepSeek V3 0324 employs advanced pre-training techniques and post-training fine-tuning tailored to specific use cases. This comprehensive approach, along with extensive compute resources, prioritizes accuracy, coherence, and reliability, making the model ideal for industries requiring long-form content generation, complex problem-solving, and enterprise-grade performance.
Benchmark
Content length
What Is Content Length?
In the context of AI models like DeepSeek, content length refers to the maximum number of tokens (words, punctuation, or symbols) that the model can process and generate in a single interaction. It defines how much input the model can understand at once and how much output it can produce. A larger context window enables an AI model to process longer inputs and incorporate a greater amount of information into each output.
How Does Content Length Impact Performance?
Content length significantly influences the model’s ability to:
- Understand Context: Longer context windows allow the model to reference more information from the input, reducing the chances of losing critical details.
- Generate Coherent Outputs: With access to the full context, the model can produce responses that maintain logical consistency across longer outputs.
- Enable Complex Reasoning: Tasks that require multi-step reasoning or tracking historical states, such as coding, technical analysis, or in-depth research, benefit greatly from extended context.
Benefits of the 160k Token Expansion
With the recent upgrade, DeepSeek-V3-0324 and DeepSeek-R1-0528 now support a 160k token limit, a significant leap from the previous 128k context window. This expansion brings several key advantages:
- Supports Ultra-Long Conversations: The extended context window allows for seamless multi-turn dialogues without losing track of prior interactions, ideal for tasks requiring long-term consistency and memory.
- Facilitates In-Depth Agent Analysis: Agents can now analyze larger datasets or perform deeper reasoning over extended periods, unlocking new possibilities for applications such as business intelligence or scientific research.
- Enables Long-Output Applications: Tasks like coding generation, document summarization, or detailed report generation can now handle larger and more complex inputs, providing comprehensive results in a single output.
By expanding the content length, DeepSeek is now better equipped to handle the demands of modern workflows, making it a powerful tool for users across diverse industries.
Accessing DeepSeek via Novita AI
Step 1: Log In and Access the Model Library
Log in to your account and click on the Model Library button.
Step 2: Choose Your Model
Browse through the available options and select the model that suits your needs.
Step 3: Start Your Free Trial
Begin your free trial to explore the capabilities of the selected model.
Step 4: Get Your API Key
To authenticate with the API, we will provide you with a new API key. Entering the “Settings“ page, you can copy the API key as indicated in the image.
Step 5: Install the API
Install API using the package manager specific to your programming language.
After installation, import the necessary libraries into your development environment. Initialize the API with your API key to start interacting with Novita AI LLM. This is an example of using chat completions API for python users to access DeepSeek R1 0528.
from openai import OpenAI
client = OpenAI(
base_url="https://api.novita.ai/v3/openai",
api_key="session_r5eB9eC3oBfDyuGn69eevTLAudts6AFa3JeHeml91oVyVe6uZnGUmAh7aO9rZxUSRocPWUBszQaiC39g4VGiAg==",
)
model = "deepseek/deepseek-r1-0528"
stream = True # or False
max_tokens = 65536
system_content = ""Be a helpful assistant""
temperature = 1
top_p = 1
min_p = 0
top_k = 50
presence_penalty = 0
frequency_penalty = 0
repetition_penalty = 1
response_format = { "type": "text" }
chat_completion_res = client.chat.completions.create(
model=model,
messages=[
{
"role": "system",
"content": system_content,
},
{
"role": "user",
"content": "Hi there!",
}
],
stream=stream,
max_tokens=max_tokens,
temperature=temperature,
top_p=top_p,
presence_penalty=presence_penalty,
frequency_penalty=frequency_penalty,
response_format=response_format,
extra_body={
"top_k": top_k,
"repetition_penalty": repetition_penalty,
"min_p": min_p
}
)
if stream:
for chunk in chat_completion_res:
print(chunk.choices[0].delta.content or "", end="")
else:
print(chat_completion_res.choices[0].message.content)
Novita AI offers both DeepSeek R1 0528 and DeepSeek V3 0324 with extended context length, allowing you to handle longer conversations and more complex projects. With DeepSeek R1 0528 priced at just $0.7/$2.5 per 1M input/output tokens and DeepSeek V3 0324 at an even more affordable $0.28/$1.14 per 1M tokens, you can access state-of-the-art AI performance without breaking the budget. For a limited time, new users can claim **$10 in free credits**to explore the upgraded DeepSeek models and various other LLM APIs on Novita AI.
Frequently Asked Questions
What is Deepseek V3 0324?
Deepseek V3 0324 is a state-of-the-art large language model from DeepSeek, designed for general conversation, comprehension, and advanced reasoning.
How does the extended context length benefit coding projects?
With extended context, you can share entire codebases, maintain context across multiple files, and work on complex refactoring tasks without losing track of dependencies and relationships between different parts of your code. This dramatically improves the AI’s ability to provide accurate, contextually relevant suggestions.
What is a good context length for an LLM?
For most applications, 32K-128K tokens provides excellent performance, while Novita AI’s upgraded 160K context length ensures you can handle even the most complex documents and extended conversations without any limitations.
About Novita AI
Novita AI is an AI cloud platform that offers developers an easy way to deploy AI models using our simple API, while also providing the affordable and reliable GPU cloud for building and scaling.
