Bedrock needs inference profile ARN instead of model name

[vladionescu]

When using [Async]AnthropicBedrock I have to provide "arn:aws:bedrock:us-east-1:0000000000:inference-profile/us.anthropic.claude-3-7-sonnet-20250219-v1:0" to model_name= instead of just anthropic.claude-3-7-sonnet-20250219-v1:0.

If I provide only the model's name, I get

Error code: 400 - {'message': 'Invocation of model ID anthropic.claude-3-7-sonnet-20250219-v1:0 with on-demand throughput isn’t supported. Retry your request with the ID or ARN of an inference profile that contains this model.'}

httpx.HTTPStatusError: Client error '400 Bad Request' for url 'https://bedrock-runtime.us-east-1.amazonaws.com/model/anthropic.claude-3-7-sonnet-20250219-v1:0/invoke'
For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/400

Docs from Anthropic or AWS don't make this obvious, and instead indicate to use the model name directly.

Maybe just needs a docs update?

Repro

from anthropic import AnthropicBedrock

client = AnthropicBedrock(
    aws_profile="bedrock",
    aws_region="us-east-1",
)

# Succeeds (replace 000000000 with your AWS account ID)
message = client.messages.create(
    model="arn:aws:bedrock:us-east-1:0000000000:inference-profile/us.anthropic.claude-3-7-sonnet-20250219-v1:0",
    max_tokens=256,
    messages=[{"role": "user", "content": "Hello, world"}],
)
print(message.content)

# Fails
message = client.messages.create(
    model="anthropic.claude-3-7-sonnet-20250219-v1:0", max_tokens=256, messages=[{"role": "user", "content": "Hello, world"}]
)
print(message.content)

[EktorTyp]

Have you tried using the Inference profile ID from the Cross-region inference? Directly without the arn.

3.5 had the same "issue" -> here