AWS Bedrock Models by Region

① bedrock-runtime — Traditional path

Anthropic Claude / Amazon Nova / Meta Llama / DeepSeek / Mistral / Cohere 等绝大多数模型走这条路径，使用 AWS SDK + InvokeModel 或 Converse。

import boto3
client = boto3.client("bedrock-runtime", region_name="us-east-1")
resp = client.converse(
    modelId="global.anthropic.claude-opus-4-8",  # ⚠️ INFERENCE_PROFILE 模型必须加前缀
    messages=[{"role":"user","content":[{"text":"Hello"}]}],
)
print(resp["output"]["message"]["content"][0]["text"])

② bedrock-mantle — OpenAI-compatible layer (NEW)

OpenAI GPT-5.x、xAI Grok 等闭源旗舰走 Mantle。用 OpenAI SDK 直接调：

from openai import OpenAI
client = OpenAI(
    api_key="<your Bedrock API key>",
    base_url="https://bedrock-mantle.us-east-2.api.aws/openai/v1",
)
# GPT-5.5 走 Responses API
r = client.responses.create(model="openai.gpt-5.5", input="hi")
# Grok / GPT-5.4 走 Chat Completions
r = client.chat.completions.create(
    model="xai.grok-4-3",
    messages=[{"role":"user","content":"hi"}],
)

③ INFERENCE_PROFILE — Cross-region routing (prefix required!)

很多新模型只支持 INFERENCE_PROFILE，调用时 modelId 必须带前缀，否则报 ValidationException：

• global.modelId — 全球路由，吞吐最高，无数据驻留限制
• us.modelId — 美区内路由，遵守 US 数据驻留
• eu.modelId — 欧区内路由
• apac.modelId — 亚太路由

表格中带 IP 标记的模型即为此类。例：global.anthropic.claude-opus-4-8

Runtime vs Mantle — at a glance