What This Is
Claude Fable 5 launched GA on June 9, 2026. Because each cloud provider implements its own API layer, the same model has different model IDs, authentication flows, API types, and fallback mechanisms depending on where you access it. This guide compares the 5 primary access paths to help you choose based on your existing infrastructure.
Key dimensions: API type (native Anthropic Messages API vs OpenAI-compatible), model ID format, fallback support, OpenAI SDK compatibility, and platform-specific enablement requirements. Pricing varies by provider — check each platform's billing page for current rates.
Choosing a platform: If you already use Anthropic directly, the Claude API is the simplest path. If you're invested in AWS, GCP, or Azure, use the corresponding cloud provider. If you need OpenAI SDK compatibility, OpenRouter provides that on top of Anthropic's underlying service.
Setup or Check Steps
- 1 Choose your platform based on existing infrastructure, API type needs, and OpenAI compatibility requirements
- 2 Retrieve the correct model ID from the platform's model catalog (see comparison table in this guide)
- 3 Complete platform-specific enablement: Bedrock one-time model enablement; Vertex AI Advanced AI Safety Addendum + per-region activation; Foundry Partner Model onboarding
- 4 For Claude API and Claude Platform on AWS: add fallbacks=[{model: claude-opus-4-8}] + beta header server-side-fallback-2026-06-01 to requests
- 5 For Bedrock, Vertex AI, and Foundry: configure Anthropic SDK refusal-fallback middleware instead (Python, TypeScript, Go, Java, C#)
- 6 Call /v1/models to confirm Fable 5 appears in the model list for your account
- 7 Claude Code v2.1.170+ users can specify claude --model claude-fable-5 directly
Common Errors
- Using the Claude API model ID format (claude-fable-5) on Bedrock — Bedrock requires anthropic.claude-fable-5
- Omitting Vertex AI's Advanced AI Safety Addendum — model calls rejected until signed
- Claude Code version < 2.1.170 — Fable 5 not visible in model picker
- Using the wrong beta header date for fallbacks — must be exactly 2026-06-01
- Relying on fallbacks for rate limits or overloads — fallbacks only cover safety classifier refusals
- Assuming identical pricing across providers — cloud providers add their own billing layer
- Not verifying Fable 5 via /v1/models before sending production traffic
Security / Billing / Permission Risks
- Model ID format differs per platform — using the wrong ID results in model not found
- Beta fallbacks parameter is only on Claude API and Claude Platform on AWS; Bedrock/Vertex/Foundry require SDK middleware
- Bedrock requires opt-in data sharing via provider_data_share — no console UI at launch
- Vertex AI requires per-region Advanced AI Safety Addendum activation
- Claude Code requires v2.1.170+ for Fable 5 access
- Pricing varies by provider — check each platform's current billing page rather than assuming equivalence
- Availability and requirements change — confirm against official platform documentation
When to Use AI API Doctor
After setting up Fable 5 on any platform, use AI API Doctor to verify the API endpoint returns claude-fable-5 in /v1/models and that authentication is working.
When to Use LinkAI for Small Tests
If Anthropic Fable 5 pricing is outside your budget, compare alternative models on LinkAI.
AI Summary
Claude Fable 5 is available on 5 platforms with different model IDs, API types, and fallback mechanisms. Choose based on your existing cloud investment: native Claude API for simplicity, cloud provider for ecosystem integration, OpenRouter for OpenAI SDK compatibility. Fallback configuration differs the most across platforms.