Claude Fable 5.1 API on Sprelay

The Sprelay model ID is claude-fable-5-1. Its live price page lists both Anthropic Messages and OpenAI-compatible endpoints, with Claude Code MAX as a compatible group.

Choose the correct API format

Client or format	Base URL or endpoint
Claude Code	Base URL https://sprelaytoken.com
Anthropic Messages request	Endpoint https://sprelaytoken.com/v1/messages
OpenAI-compatible client	Base URL https://sprelaytoken.com/v1
Model ID in either format	claude-fable-5-1

Checked in the public catalog on 11 October 2026. Availability and feature compatibility need to be verified with your own key. No authenticated inference test is claimed here.

Pricing and an example workload

For Claude Code MAX, the live page showed these dated example rates in USD per 1 million tokens:

Input	Output	Cached input	Cache write
$4.00	$20.00	$0.10	$1.50

For 10,000 uncached input tokens and 2,000 output tokens, the illustrated cost is 0.01 × $4 + 0.002 × $20 = $0.08.

Cache writes and additional agent calls are excluded. Cache-read pricing applies only to eligible usage recorded as a cache hit. The current pricing page and usage record take precedence; read billing.

Configure Claude Code

1. Create a dedicated key, choose a compatible Claude Code MAX group and set a small quota.
2. Follow Claude Code configuration for your operating system.
3. Set ANTHROPIC_MODEL to claude-fable-5-1 and keep ANTHROPIC_BASE_URL at the site root.
4. Start a new Claude Code process from that terminal and ask it to explain a small file.
5. Check usage logs for the correct key, model, status and charge.

For a direct protocol test, use the Claude Messages request with this model ID. For an OpenAI-compatible desktop client, see Cherry Studio setup.

Resolve common failures

• A duplicated /v1/v1/messages path can cause 404. Claude Code's base URL must not include /v1.
• An Anthropic login flow can indicate that the new process did not inherit the intended gateway configuration. Check for conflicting credential settings.
• A model access error requires checking the exact ID and selected key group.
• Unexpected spend requires inspecting all calls in an agent task, context size and cache records.

See API errors. For support, share a timestamp and redacted request ID, never the full token.

Related models

GPT 6.1 Sol · GPT 6 Astra