AWS Bedrock

Language

Examples use Python. Beakpoint works with any language that OpenTelemetry supports.

Required AWS Bedrock Attributes

The following cost calculation tags are required for traces originating from AWS Bedrock API calls.

Always Required Fields

These fields are always required for Beakpoint to calculate Bedrock costs:

Attribute Name	Example Value	Allowed Values
gen_ai.system	aws.bedrock	aws.bedrock (must be this exact value)
gen_ai.request.model	anthropic.claude-sonnet-4-20250514-v1:0	Any valid Bedrock model ID
gen_ai.usage.input_tokens	512	Non-negative integer
gen_ai.usage.output_tokens	128	Non-negative integer

Optional Enrichment Attributes

These fields are optional but improve cost accuracy when provided:

Attribute Name	Example Value	Description
gen_ai.response.model	anthropic.claude-sonnet-4-20250514-v1:0	The exact model version returned in the response. Used as a fallback if gen_ai.request.model is absent.
gen_ai.provider.name	aws.bedrock	OTel semconv v1.37.0+ standardized attribute (replaces gen_ai.system). Use this for new instrumentation; Beakpoint accepts both, but prefers gen_ai.provider.name when present.
cloud.provider	aws	Provider identification.

Model ID Handling

Bedrock model IDs follow the format [region.]provider.model-name[-date]-v{version}[:{variant}]. Beakpoint automatically parses these to extract the underlying provider and model for pricing lookup. Pass the model ID exactly as Bedrock provides it — no normalization is needed on your part.

For details on how model IDs are parsed, see GenAI Cost Tracking — AWS Bedrock Model ID Parsing.

tip

Models from providers that Beakpoint already prices directly (Anthropic, OpenAI, Mistral, Meta, Cohere) use the same pricing whether called through Bedrock or through the provider's direct API. Bedrock does not add a per-token surcharge for these models.

Bedrock-Exclusive Models

The following models are available only through AWS Bedrock. Prices represent cost per 1 million input or output tokens, respectively.

Provider	Model	Display Name	Input ($/M)	Output ($/M)
amazon	nova-pro	Amazon Nova Pro	$0.80	$3.20
amazon	nova-lite	Amazon Nova Lite	$0.06	$0.24
amazon	nova-micro	Amazon Nova Micro	$0.035	$0.14
meta	llama4-maverick-17b-128e-instruct	Llama 4 Maverick	$0.20	$0.60
meta	llama4-scout-17b-16e-instruct	Llama 4 Scout	$0.15	$0.30
mistral	mistral-large-2411	Mistral Large	$2.00	$6.00
mistral	mistral-small-2501	Mistral Small	$0.10	$0.30
cohere	command-r-plus	Command R+	$2.50	$10.00
cohere	command-r	Command R	$0.15	$0.60

note

Models from other providers (Anthropic Claude, OpenAI GPT) are also available on Bedrock and are priced using the same rates as the direct API. See the Anthropic Claude and OpenAI reference pages for those prices.

Python Example

The quickest way to emit the required attributes is with the opentelemetry-instrumentation-botocore package, which includes Bedrock extensions:

pip install opentelemetry-instrumentation-botocore

from opentelemetry.instrumentation.botocore import BotocoreInstrumentor
import boto3

# Instrument before creating the client. For full TracerProvider setup including
# _GenAiSystemProcessor, see the Track LLM Costs guide (referenced at the end).
BotocoreInstrumentor().instrument()

client = boto3.client("bedrock-runtime")

response = client.invoke_model(
    modelId="anthropic.claude-sonnet-4-20250514-v1:0",
    contentType="application/json",
    accept="application/json",
    body='{"messages": [{"role": "user", "content": "Hello!"}], "max_tokens": 4096, "anthropic_version": "bedrock-2023-05-31"}',
)

The instrumentation automatically sets gen_ai.system, gen_ai.request.model, gen_ai.usage.input_tokens, and gen_ai.usage.output_tokens on each span.

For full setup instructions, including how to configure the OpenTelemetry exporter for Beakpoint, see the Track LLM Costs guide.