|
|
As of January 30, the Deepseek-R1 models have become available in Amazon Bedrock through the Amazon Bedrock Marketplace and Amazon Bedrock Custom Model Import. Since then, thousands of customers have deployed these models in Amazon Bedrock. Customers appreciate robust railing and complex tools for safe deployment. Today, it is even easier to use Deepseek in Amazon Bedrock through an extended range of options, including a new solution without a server.
The fully managed Deepseek-R1 is now generally available in Amazon Bedrock. Amazon Web Services (AWS) is the first cloud service provider (CSP) to supply Deepseek-R1 as a fully managed, generally available model. You can get innovations and provide a tangible business value with Deepseek at AWS without having to drive an additional infrastructure. You can power your AI generative applications with Deepseek-R1 capabilities using a single API in the fully managed Amazon Bedrock service and gain the advantage of its extensive features and tools.
According to Deepseek, their model is publicly available on the basis of a MIT license and offers strong capabilities in justification, coding and understanding natural language. Diploma diploma intelligent decision -making decisions, software development, mathematical problem solving, scientific analysis, data and understanding of knowledge management systems.
As with all AI solutions, give careful consideration of personal data protection requirements when implementing in your production, check output distortion and monitor your results. When implementing publicly available models such as Deepseek-R1, consider the following:
- Data security -You have access to business security, monitoring and controlling the cost of Amazon Bedrock, which are necessary to deploy AI responsibly on a scale, all while maintaining full control over your data. User inputs and model outputs are not shared with model providers. You can use these key safety features by default, including data encryption at rest and transit, fine-grained access controls, secure connection options, and download different conformation certifications when communicating with the Deepseek-R1 model in Amazon Bedrock.
- AI responsible – You can implement guarantees adapted to your requirements for your applications and responsible for the police with Amazon railing. This included key content filtering properties, sensitive information filtering, and customizable safety controls to prevent hallucinations by contextual grounding and automated thinking checks. This means that you can control the interaction between users and Deeepseek-R1 in the subsoil with a defined set of police filtering of undesirable and harmful content in your generative AI applications.
- Evaluation of the model -You can evaluate and compare models to identify the optimal model for your use, including Deepseek-R1, in several steps through automatic or human evaluation using tools for evaluating the Amazon Bedrock. You can choose automatic evaluation with predefined metrics such as accuracy, robustness and toxicity. Alternatively, you can choose workflows for human evaluation for subjective or your own metrics such as Livage, Style and Alignment with a branded voice. Providing a model of the built -in built -in curatorial data set or bringing to your own data sets.
We strongly recommend the integration of Amazon Bedrock Guardrails and using the Amazon Bedrock features with your Deepseek-R1 model to add robust protection for your generative AI applications. If you want to learn more, visit your Deepseek’s deployment using Amazon Bedrock Guardrails and evaluate the performance of Amazon Bedrock resources.
Start with Deepseek-R1 in Amazon Bedrock
If you are a newcomer in using Deepseek-R1 models, go to the Amazon Bedrock console, choose Access to the model under Configuration In the left navigation pane. To access the fully managed Deeepseek-R1 model, ask for access Deepseek-R1 in Deepseek. You will then have grants access to the Amazon Bedrock model.
Further to test the Deepseek-R1 model in Amazon Bedrock, choose Cat/text under Playground On the left pane. Then he chooses Select the model At the top left and select Deepseek as a category of a Deepseek-R1 as a model. Then he chooses App.
By means of a selected Deepseek-R1 Model, I will start the following rapid examination:
A family has $5,000 to save for their vacation next year. They can place the money in a savings account earning 2% interest annually or in a certificate of deposit earning 4% interest annually but with no access to the funds until the vacation. If they need $1,000 for emergency expenses during the year, how should they divide their money between the two options to maximize their vacation fund?
This challenge requires a complex chain of thinking and brings very accurate reasoning results.
To read more about recommendations for challenges, Fer on the Readme model Deepseek-R1 in your Github Rekot.
By selecting View the API requestYou can also approach examples of using the model code in the AWS (AWS CLI) and AWS SDK command line. You can use us.deepseek.r1-v1:0 Like a model ID.
Here is a sample of the AWS clinic:
aws bedrock-runtime invoke-model \
--model-id us.deepseek-r1-v1:0 \
--body "{\"messages\":({\"role\":\"user\",\"content\":({\"type\":\"text\",\"text\":\"(n\"})}),max_tokens\":2000,\"temperature\":0.6,\"top_k\":250,\"top_p\":0.9,\"stop_sequences\":(\"\\n\\nHuman:\")}" \
--cli-binary-format raw-in-base64-out \
--region us-west-2 \
invoke-model-output.txtThe model supports both InvokeModel and Converse API. The following Python Examleles code shows how to send a text message to DeepSeek-R1 using Amazon Bedrock Converse API to generate text.
import boto3
from botocore.exceptions import ClientError
# Create a Bedrock Runtime client in the AWS Region you want to use.
client = boto3.client("bedrock-runtime", region_name="us-west-2")
# Set the model ID, e.g., Llama 3 8b Instruct.
model_id = "us.deepseek.r1-v1:0"
# Start a conversation with the user message.
user_message = "Describe the purpose of a 'hello world' program in one line."
conversation = (
{
"role": "user",
"content": ({"text": user_message}),
}
)
try:
# Send the message to the model, using a basic inference configuration.
response = client.converse(
modelId=model_id,
messages=conversation,
inferenceConfig={"maxTokens": 2000, "temperature": 0.6, "topP": 0.9},
)
# Extract and print the response text.
response_text = response("output")("message")("content")(0)("text")
print(response_text)
except (ClientError, Exception) as e:
print(f"ERROR: Can't invoke '{model_id}'. Reason: {e}")
exit(1)To enable the Amazon Bedrock railing on the Deepseek-R1 model, select Railing under Warranty In the left navigation pane and create a railing with configuration as many filters as you need. For example, if you filter for the word “politics”, your railing will recognize this word in the challenge and show you a blocked message.
You can try railing with different inputs to assess the performance of the railing. You can specify the warranty by setting the topics, words filters, sensitive information filters and blocked messages until it matches your needs.
If you want to learn more about Amazon Bedrock Guardrails, visit the stopping of malicious content in models using Amazon Bedrock Guardrails in AWS documentation or other deep submersible blog posts on Amazon Bedrock Guardrails on AWS Machine Learning blog.
Here is a demo passage that emphasizes how you can use the fully managed Deepseek-R1 model in Amazon Bedrock:
Now available
Deepseek-R1 is now Avaidable fully managed in Amazon Bedrock at the US East (N. Virginia), US East (Ohio) and USA West (Oregon) AWS through Crossregion Inference. For future updates, see the entire Region list. If you want to know more, check out the Amazon Bedrock page and Amazon Bedrock.
Try Deepseek-R1 today in Amazon Bedrock Console and send AWS Re: Post for Amazon Bedrock or through the usual AWS support contacts.
– Channels
Updated March 10, 2025 – Fixed images of the model selection screen and model ID.