Bedrock model ID: complete reference for AWS AI

Every API call to Amazon Bedrock requires a Bedrock model ID — a string like anthropic.claude-3-5-sonnet-20241022-v2:0 that tells AWS exactly which foundation model to invoke. Get it wrong and you get a ValidationException. Get it right and you have access to models from Anthropic, Meta, Mistral, Cohere, Amazon, and others through a single API.

As of early 2025, Bedrock offers access to over 40 foundation models across 7 model providers. CodeWords connects to these models through its LLM integrations, giving you multi-provider AI access without managing AWS credentials yourself.

Unlike generic AI automation posts, this guide shows real CodeWords workflows — not just theory.

TL;DR

• Bedrock model IDs follow the format provider.model-name-version:iteration
• Model IDs differ by region — check us-east-1 and us-west-2 first for widest selection
• CodeWords abstracts away Bedrock model IDs by providing direct LLM access without API key setup

How are Bedrock model IDs structured?

Format: {provider}.{model-name}-{version}:{iteration}

• Provider prefix: anthropic, meta, mistral, cohere, amazon, ai21, stability
• Model name: The specific model
• Version: Date-based or numeric version
• Iteration: Numeric suffix (usually 0 or 1)

Current Bedrock model IDs

Anthropic Claude models

• anthropic.claude-3-5-sonnet-20241022-v2:0 — Best balance of speed and quality
• anthropic.claude-3-5-haiku-20241022-v1:0 — Fastest and cheapest
• anthropic.claude-3-opus-20240229-v1:0 — Highest capability

Meta Llama models

• meta.llama3-2-90b-instruct-v1:0
• meta.llama3-2-11b-instruct-v1:0 (multimodal)
• meta.llama3-1-405b-instruct-v1:0

Mistral models

• mistral.mistral-large-2411-v1:0
• mistral.mistral-small-2402-v1:0

Amazon Titan models

• amazon.titan-text-premier-v1:0
• amazon.titan-embed-text-v2:0 (for RAG/search)

Cohere models

• cohere.command-r-plus-v1:0 — Best for RAG
• cohere.embed-english-v3

Using Bedrock model IDs in code

import boto3, json

bedrock = boto3.client("bedrock-runtime", region_name="us-east-1")

def invoke_claude(prompt, model_id="anthropic.claude-3-5-sonnet-20241022-v2:0"):
    response = bedrock.invoke_model(
        modelId=model_id,
        body=json.dumps({
            "anthropic_version": "bedrock-2023-05-31",
            "max_tokens": 1024,
            "messages": [{"role": "user", "content": prompt}]
        }),
        contentType="application/json",
    )
    return json.loads(response["body"].read())["content"][0]["text"]

Which model ID should you choose?

• General reasoning: Claude 3.5 Sonnet v2
• High-volume tasks: Claude 3.5 Haiku or Mistral Small
• Open-source: Llama 3.2 90B
• RAG: Cohere Command R+ with Titan Embeddings v2

In CodeWords, you get direct access to OpenAI, Anthropic, and Google Gemini without configuring AWS credentials.

FAQs

Why does my model ID return a ValidationException? Check for typos, model access not enabled, or region unavailability.

Are IDs the same across regions? Strings are the same but availability varies by region.

How do I list all available models? Use boto3.client("bedrock").list_foundation_models().

Explore multi-model AI workflows in CodeWords