Selecting the right
text-based LLM
for your use case
Maximize the business value of LLMs with
these insights and strategies
Generative artificial intelligence (gen AI) applications powered
by large language models (LLMs) have the potential to transform
every industry. Staying up to speed with LLMs is challenging as
the technology continues to evolve at lightning speed.
Read on to discover best practices that will help you navigate the
LLM landscape with confidence.
What can text-based LLMs do for you?
LLMs are trained on trillions of words across many natural language tasks.
When supported by the right gen AI infrastructure, they can carry out functions
in a conversational manner. These functions include:
Engaging in Understanding, Answering Summarizing Providing
interactive learning, and questions dialogues and suggestions
conversations generating text documents
LLMs are powering applications across multiple industries,
including healthcare and life sciences, media and entertainment,
financial services, and more.
Key factors to consider
for text-based LLMs
The list of available LLMs is growing fast as technology evolves. Understanding the
factors that shape LLM options can help you with selection, gaining a competitive
advantage, and increasing the business value of your gen AI investments.
Model
capabilities
Assess the model’s ability to perform specific
tasks, such as text generation, translation,
summarization, and code generation.
Model size
and complexity
Larger models often offer better performance but
require more computational resources. Evaluate your
specific needs to determine the optimal size.
Model
customization
Determine whether the model can be used
directly for your tasks or if it requires
fine-tuning your specific domain data
Cost and
to achieve optimal performance.
accessibility
Factor in the cost of using the model, including
API fees or licensing costs. Consider
open source options for more flexibility.
Ethical
considerations
Be mindful of potential biases in the model’s
outputs and take steps to mitigate them.
Ensure the model aligns with ethical guidelines
and principles.
Find the right text-based
LLM for your use case
Beyond the factors that shape LLMs, selecting the right text-based LLM for your use case can increase
the accuracy and quality of your inputs, improve performance, and drive greater cost-efficiency.
TA SK S
TE X T CL A SSIFIC ATIO N
TE X T G E N E R ATIO N
Q&A
T R A N SL ATIO N
C O N V E R S ATIO N A L A I
S U M M A R IZ ATIO N
D O C U M E N T U N D E R STA N DIN G
S O F T W A R E D E V ELO P M E N T
LIFE S CIE N CES
Interested in other LLM use cases? Go beyond text-based LLMs with the broadest selection
of multi-modal options on Amazon SageMaker AI and Amazon Bedrock, including the new
Amazon Nova foundation models (FMs). Learn more about leveraging your customized
models across AI21 Labs, Anthropic, Luma AI, Meta, Mistral, Stability.AI, and more.
Learn more ›
Choose the right infrastructure to
optimize performance and costs for LLMs
Build with specialized AI infrastructure that delivers the performance you need while reducing costs.
1 2
Rightsize Choose the optimal
your model infrastructure
You may not need the largest model. Explore purpose-built infrastructure solutions
Pick the right type and size model that are uniquely designed from the ground up
depending on your use case. to accelerate innovation, enhance security, and
improve performance while lowering costs.
AWS offers infrastructure and services designed to help
you get the most performance out of your LLMs while
optimizing your costs:
ACCELER ATE D CO M P U TIN G
From the highest-performance NVIDIA GPU-based Amazon Elastic Compute
Cloud (Amazon EC2) to continued investments in our purpose-built machine
learning (ML) accelerators AWS Trainium and AWS Inferentia, AWS delivers the
best price performance for training and deploying gen AI models at scale.
A M A ZO N SAG E M A K ER A I
Amazon SageMaker AI is a fully managed service that brings together a broad
set of tools to enable high-performance, low-cost ML for any use case. With
SageMaker AI, you can build, train, and deploy ML models at scale using tools,
such as notebooks, debuggers, profilers, pipelines, machine learning operations
(MLOps), and more—all in one integrated development environment (IDE).
N E T W O R K IN G
Purpose-built to meet the performance demands for gen AI, Amazon Web
Services (AWS) provides high-throughput and low-latency networking that
includes Elastic Fabric Adapter (EFA) and Amazon EC2 UltraClusters.
STO R AG E
Accelerate compute workloads with Amazon FSx for Lustre, which provides
sub-millisecond latencies, up to hundreds of gigabytes per second (GBps) of
throughout, and millions of input/output operations per second (IOPS) while
quickly accessing and processing your datasets on Amazon Simple Storage
Service (Amazon S3).
SECU RIT Y
Our accelerated computing Amazon EC2 instances and networking are built on a
foundation of the AWS Nitro System, which has been validated by independent
cybersecurity firm NCC Group. The level of security protection offered is so critical
that we’ve added it to our AWS Service Terms to provide additional assurance to
all of our customers.
AWS Trainium2–based Amazon AWS Inferentia2–based
EC2 Trn2 instances deliver Amazon EC2 Inf2 instances deliver
U P TO U P TO
30%
better price performance than current
4 0%
lower cost per inference than
generation GPU-based EC2 instances comparable EC2 instances
Unleash the power of
generative AI LLMs
Optimize performance and costs for LLM deployment
While LLMs hold the potential to transform your business and give it a competitive
edge, building, training, and deploying them requires an unprecedented level
of infrastructure resources. To succeed, you need an infrastructure strategy that
delivers the right processing power without compromising on cost or performance;
low-latency, high-throughput networking; storage solutions that help accelerate
and cost-optimize your compute; and a deep set of cloud services and partners.
Empower your organization with LLMs—start your journey with AWS today.
Get started with AWS AI infrastructure ›
© 2025, Amazon Web Services, Inc. or its affiliates. All rights reserved.
