Claude Fable 5.2 is now available on Amazon Bedrock.
This page has been translated by machine translation. View original
Introduction
On November 9, 2026, Claude Fable 5.2, Anthropic's generally available highest-performance model for technical work, became available on Amazon Bedrock.
This article introduces the results of a test calling the Converse API with a Global inference profile specified, along with a check of the pricing.
Pricing
I checked the Global input and output unit prices per 1 million tokens on the Bedrock pricing page.
| Model | Input | Output |
|---|---|---|
| Fable 5.2 | $8.00 | $40.00 |
| Mythos 5.2 | $8.00 | $40.00 |
| Fable 5.1 | $10.00 | $50.00 |
| Opus 5.5 | $4.00 | $20.00 |
Compared to Fable 5.1, Fable 5.2 is $2 cheaper for input and $10 cheaper for output, a 20% reduction in both cases.
When using US Geo, the unit prices (us-east-1) are 10% higher than Global: $8.80 for input and $44.00 for output.
Model IDs and Available Regions
In Bedrock, it is offered through US Geo and Global inference profiles. The IDs listed on the model card are as follows.
| Offering type | Model ID |
|---|---|
| Base | anthropic.claude-fable-5-2 |
| Geo (US) | us.anthropic.claude-fable-5-2 |
| Global | global.anthropic.claude-fable-5-2 |
In the Tokyo region (ap-northeast-1) of the test account, I ran list-inference-profiles.
aws bedrock list-inference-profiles --region ap-northeast-1 --type-equals SYSTEM_DEFINED --max-results 1000 --output json | jq '{inferenceProfileSummaries: [.inferenceProfileSummaries[] | select(.inferenceProfileId|test("fable-5"))]}'
I excerpted the main items for Fable 5.2 from the output.
{
"inferenceProfileName": "Global Anthropic Claude Fable 5.2",
"description": "Routes requests to Claude Fable 5.2 globally across all supported AWS Regions.",
"inferenceProfileId": "global.anthropic.claude-fable-5-2",
"status": "ACTIVE",
"type": "SYSTEM_DEFINED"
}
Changes from Fable 5.1
Based on the model cards for Fable 5.0, 5.1, and 5.2, I summarized the differences in a table.
| Item | 5.0 | 5.1 | 5.2 |
|---|---|---|---|
| Release date | 2026-06-09 | 2026-09-01 | 2026-11-09 |
| Knowledge cutoff | January 2026 | June 2026 | June 2026 |
| bedrock-mantle mentioned on the model card | Yes | Yes | No (bedrock-runtime only) |
| Blocked areas | Cyber, biology (dual-use) | Cyber, life sciences (dual-use) | High-risk cyber activities, most biological research tasks |
The points that are the same in 5.1 and 5.2 are as follows.
- Context is 1M tokens, and maximum output is 128K
- temperature is fixed at 1.0 or unspecified
- Adaptive thinking is always ON, effort is low / medium / high / xhigh / max, and the default is high
- Prompt caching has a minimum of 512 tokens, a maximum of 4 checkpoints, and a TTL of 5 minutes or 1 hour
Data Retention Mode
Anthropic designates Claude Fable 5.2 as a Covered Model, and using it requires setting the data retention mode to aws_review.
aws bedrock put-account-data-retention --region ap-northeast-1 --mode aws_review
For details, see the Fable 5.1 article.
Calling with the Converse API
From the Tokyo region, I called the Converse API once with the Global profile ID specified.
aws bedrock-runtime converse --region ap-northeast-1 --model-id global.anthropic.claude-fable-5-2 --messages '[{"role":"user","content":[{"text":"こんにちは。自己紹介を一文で。"}]}]' --inference-config '{"maxTokens":100}' --output json
The response is as follows. The signature value in reasoningText has been replaced with <signature>.
{
"output": {
"message": {
"role": "assistant",
"content": [
{
"reasoningContent": {
"reasoningText": {
"text": "",
"signature": "<signature>"
}
}
},
{
"text": "こんにちは、私はAnthropicが開発したAIアシスタントのClaudeです。"
}
]
}
},
"stopReason": "end_turn",
"usage": {
"inputTokens": 24,
"outputTokens": 76,
"totalTokens": 100,
"cacheReadInputTokens": 0,
"cacheWriteInputTokens": 0
},
"metrics": {
"latencyMs": 2516
}
}
In this call, the first element of content was reasoningContent, and the response body text was the second element.
Summary
I confirmed that Claude Fable 5.2 is now available on Amazon Bedrock.
The pricing of Fable 5.2 is 20% cheaper than Fable 5.1 for both input and output. Please give it a try as a frontier model offered by Anthropic.
In addition, Claude Mythos 5.2, which is built on the same foundation model as Claude Fable 5.2 and has cyber and bio capabilities, has also started to be offered on Bedrock. It is covered by Anthropic's vetting program, and if Mythos 5.2 is available to you, I recommend evaluating it as well.
