Fable 5 API Quickstart

Three-minute setup

Full docs: docs.omniakey.com — Anthropic protocol on the bare host, OpenAI protocol at /v1/chat/completions; the same key also covers GPT-5.5 and Gemini 3.1.

Python

from anthropic import Anthropic

client = Anthropic(
    api_key="YOUR_API_KEY",
    base_url="https://api.omniakey.com",  # drop this line for the official endpoint
)

message = client.messages.create(
    model="claude-fable-5",
    max_tokens=2048,
    messages=[{"role": "user", "content": "Introduce Fable 5 in one sentence"}],
)
print(message.content[0].text)

Node.js

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  apiKey: "YOUR_API_KEY",
  baseURL: "https://api.omniakey.com", // drop for the official endpoint
});

const msg = await client.messages.create({
  model: "claude-fable-5",
  max_tokens: 2048,
  messages: [{ role: "user", content: "Hello, Fable 5" }],
});
console.log(msg.content[0].text);

cURL

curl https://api.omniakey.com/v1/messages \
  -H "x-api-key: $YOUR_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-fable-5",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello"}]
  }'

How to get a Fable 5 API key

Two routes, by audience:

Official: sign up at Claude Console, add a card, generate a key on the API Keys page. OmniaKey: sign up → top up (balance never expires; unused balance refundable within 30 days) → generate a key, then point base_url at it in the three-minute setup above.

Pricing at a glance

That's exactly 2× Opus 4.8 ($5/$25). Run your numbers with the cost guide.

Fallback billing: the gotcha to know

• Log the fallback flag — monitor your downgrade rate, especially for security or bio products.
• Budget at a blended rate for high-sensitivity domains.
• Handle quality fallbacks — flows that depend on the 1M context should detect downgrades and decide: retry or accept.

Background on the guardrails — and why the unguarded Claude Mythos 5 isn't public — on our sister site.

Engineering notes

• Always stream. Time-to-first-token can reach ~80 seconds; non-streaming calls will trip default client timeouts. Set HTTP timeouts to the 10-minute range.
• Lean on prompt caching. Cache hits cost $1/M — long system prompts, RAG context, and few-shot examples belong in cache. It's the biggest cost lever.
• Reserve it for hard work. For short, well-defined tasks, swap the model string to Opus 4.8 or Sonnet 4.6 — same code.
• Compliance: Fable 5 requests carry a mandatory 30-day retention window (safety monitoring; not used for training).