How Amazon Bedrock’s Three New Service Tiers Let You Balance Performance and Cost

Amazon Bedrock now offers three distinct service tiers—Priority, Standard, and Flex—designed to align with different AI workload characteristics. Priority processes requests first, targeting latency‑critical applications such as real‑time chat assistants and language translation, but it carries a higher price. Standard provides stable performance at regular pricing for routine tasks like content generation, text analysis, and document processing. Flex targets workloads that can tolerate higher latency, offering the lowest price and is suited for model evaluation, summarization, and multi‑step agent workflows.

The article explains that many customers struggle to balance performance and cost when running AI workloads. By selecting the appropriate tier, users can optimize spend: for example, a customer‑facing chat assistant would use Priority for the fastest response, while a summarization job could use Flex to reduce expenses while still meeting reliability needs.

For models that support Priority, the latency can be reduced by up to 25% compared to Standard, resulting in higher tokens‑per‑second output. Users are encouraged to consult the latest model‑tier compatibility list in the official Bedrock documentation.

A practical selection framework is provided: first identify which workloads require instant response versus those that can accept slower processing, then optionally split a portion of traffic across tiers to benchmark performance and cost. The Amazon Pricing Calculator can be used to estimate fees for each tier based on expected workload, helping to create a realistic budget.

Monitoring tools such as the AWS Service Quotas console, Bedrock model‑call logging, and Amazon CloudWatch metrics allow users to track token usage and tier‑specific performance, supplying data for informed tier‑selection decisions.

Example Python code demonstrates how to invoke a Bedrock model with the service_tier parameter set to "priority", "default", or "flex" using the OpenAI‑compatible API endpoint:

from openai import OpenAI
client = OpenAI(
    base_url="https://bedrock-runtime.us-west-2.amazonaws.com/openai/v1",
    api_key="$AWS_BEARER_TOKEN_BEDROCK"  # Replace with actual API key
)
completion = client.chat.completions.create(
    model="openai.gpt-oss-20b-1:0",
    messages=[
        {"role": "developer", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"}
    ],
    service_tier="priority"  # options: "priority" | "default" | "flex"
)
print(completion.choices[0].message)

For further details, readers should refer to the Amazon Bedrock service‑tier documentation and the general Bedrock user guide.