|
New AI models from META, LLAM 4 Scout 17B and LLAM 4 Maverick 17B are now available as a fully managed option without server in Amazon Bedrock. These new Endowment Models (FMS) provide native multimodal capabilities with early fusion technology that you can use to accurately display image and extended context processing in your applications.
Llama 4 uses innovative architecture of a mixture (MOE), which provides increased performance within the tasks and understanding of image while optimizing costs and speed. This architectural approach allows Llam 4 to offer improved performance at lower costs compared to Llamo 3 with extended language support for global applications.
The models were on Amazon Sagemaker Jumsstart Alerady Avaiaable and now you can use them in Amazon Bedrock to streamline and scaling generative applications with the security and privacy of business level.
LLAM 4 MAVERICK 17B – Native multimodal model with 128 experts and a total of 400 billion parameters. It excels in understanding image and text, so it follows for versatile assistants and chat applications. The model supports a context window of 1 million tokens, giving you the flexibility for processing lengthy documents and complex inputs.
LLAM 4 Scout 17b -General purple multimodal model with 16 experts, 17 billion active parameters and 109 billion total parameters that provide supernior performance compared to all previous LLAM models. Amazon Bedrock currently supports a 3.5 million context window context window for Scout Llama 4 with plans to expand in the near future.
Use boxes for Lama 4 models
You can use the advanced Llama 4 capabilities for a wide range of cases across industries:
Business applications -Intelligent assembly that can consider across tools and workflows, multimodal processes inputs and provides high -quality obligations for business applications.
Multilingual assistants -Creat chat applications that understand images and provide high -quality duties across different languages, which makes them accessible to a global audience.
Code and Documentation of Intelligence – It develops that it can understand code, extract structural data from documents and provides a thorough analysis across large volumes of text and code.
Customer support – Improve support systems using the capability of image analysis and allow a more efficient solution to the problem when customers share images or photos.
Happy creature – Generate creative content in multiple languages with the ability to understand and respond to visual inputs.
Research – Create research applications that can integrate and analyze multimodal data and provide knowledge across text and images.
Using LLAM 4 models in Amazon Bedrock
To make these new models without a server without a server in Amazon Bedrock, I have to ask for access first. I will choose in the Amazon Bedrock console Access to the model From the navigation pane to switch LLAM 4 MAVERICK 17B and LLAM 4 Scout 17b Models.
The LLAMA 4 models can be easily integrated into Amazon Bedrock Converse API applications that provide a uniform AI conversational interaction interface.
Here is an example of how to use the AWS SDK for Python (Boto3) with Llamou 4 Maverick for a multimodal conversation:
import boto3
import json
import os
AWS_REGION = "us-west-2"
MODEL_ID = "us.meta.llama4-maverick-17b-instruct-v1:0"
IMAGE_PATH = "image.jpg"
def get_file_extension(filename: str) -> str:
"""Get the file extension."""
extension = os.path.splitext(filename)(1).lower()(1:) or 'txt'
if extension == 'jpg':
extension = 'jpeg'
return extension
def read_file(file_path: str) -> bytes:
"""Read a file in binary mode."""
try:
with open(file_path, 'rb') as file:
return file.read()
except Exception as e:
raise Exception(f"Error reading file {file_path}: {str(e)}")
bedrock_runtime = boto3.client(
service_name="bedrock-runtime",
region_name=AWS_REGION
)
request_body = {
"messages": (
{
"role": "user",
"content": (
{
"text": "What can you tell me about this image?"
},
{
"image": {
"format": get_file_extension(IMAGE_PATH),
"source": {"bytes": read_file(IMAGE_PATH)},
}
},
),
}
)
}
response = bedrock_runtime.converse(
modelId=MODEL_ID,
messages=request_body("messages")
)
print(response("output")("message")("content")(-1)("text"))This example shows how to send text and image inputs to the model and receive a conversational response. The Converse API abstract leaves the complexity of working with different input model formats and provides a clear interface across the Amazon Bedrock models.
For multiple interactive use of boxes, you can also use the API Converse streaming capabilities:
response_stream = bedrock_runtime.converse_stream(
modelId=MODEL_ID,
messages=request_body('messages')
)
stream = response_stream.get('stream')
if stream:
for event in stream:
if 'messageStart' in event:
print(f"\nRole: {event('messageStart')('role')}")
if 'contentBlockDelta' in event:
print(event('contentBlockDelta')('delta')('text'), end="")
if 'messageStop' in event:
print(f"\nStop reason: {event('messageStop')('stopReason')}")
if 'metadata' in event:
metadata = event('metadata')
if 'usage' in metadata:
print(f"Usage: {json.dumps(metadata('usage'), indent=4)}")
if 'metrics' in metadata:
print(f"Metrics: {json.dumps(metadata('metrics'), indent=4)}")
When streaming, your application can provide a more sensitive experience by displaying the model outputs as they are generated.
What to know
Today, the LLAM 4 is available with a fully managed experience without servers in Amazon Bedrock on US East (N. Virginia) and the US West (Oregon) AWS regions. You can also access LLAM 4 on US East (Ohio) via Cross-Region Inference.
As usual at Amazon Bedrock you pay for what you use. For more information, see Amazon Bedrock.
These models support 12 languages for text (English, French, German, Hindi, Italian, Portuguese, Spanish, Thai, Arabic, Indonesian, Tagalog and Vietnamese) and English in image processing.
If you want to start using these new models, visit the META LLAMA Models in the Amazon Bedrock user manual. You can also explore how our architectural communities use Amazon Bedrock in their solutions in the generative section AI of our community.
– Danilo
How’s the Blog of news? Take this 1 minute survey!
(This survey is hosted by an external company. AWS processes your information as described in the AWS Privacy Notice. AWS will own data collected via this survey and will not share information with dew survey.)