Claude Fable 5 API: Access and Usage Guide

Claude Fable 5 is the first member of Anthropic's new Mythos class — a flagship built for autonomous workflows and large-scale coding, with a 1M-token context window, 128K-token output, and reasoning that continues throughout a task instead of happening once up front. The question developers keep asking is simple: how do you get API access, and how do you use it well in production?

This guide covers what Fable 5 actually is, when to choose it over Sonnet 5 or Opus 4.8, how to send your first request in Python and curl, and a routing strategy that keeps costs under control. Every example runs on Onysoft AI Gateway, which puts Fable 5 and 708+ other models behind a single OpenAI-compatible API — whether you are calling from Turkey or anywhere else in the world.

What Is Claude Fable 5? Meet the Mythos Class

Claude Fable 5 is the first model in Anthropic's new Mythos class, positioned above the familiar Opus / Sonnet / Haiku hierarchy. It is not an incremental version bump; it is a separate category built for long-horizon, autonomous work.

What stands out for developers:

• 1M-token context window: keep an entire monorepo, hundreds of pages of documentation, or a multi-day agent session in a single context.
• 128K-token output: long reports, sweeping refactors, and multi-file code generation finish in one response.
• Continuous reasoning: the model maintains and updates its plan throughout a task instead of thinking once up front, which keeps multi-step work coherent.
• Autonomous work and coding focus: optimized for agent architectures, long-running background jobs, and complex engineering tasks.

On Onysoft it ships as anthropic/claude-fable-5; you can browse the full lineup on the Claude models page.

Fable 5 vs Sonnet 5 vs Opus 4.8: Which One, When?

All three models live behind the same API, so switching is a one-line model change. The right question is not "which is best" but "which model does this job belong to":

• Claude Fable 5: multi-step refactors across huge codebases, autonomous agent runs that last hours, analyses that genuinely need the 1M context. The hardest ten percent of your workload.
• Claude Opus 4.8: tasks that need deep reasoning but not Mythos scale — architecture reviews, complex technical analysis, hard one-shot problems.
• Claude Sonnet 5: everyday coding assistance, production agent workloads, code review; the sweet spot of speed and quality.
• Claude Haiku 4.5: classification, summarization, and high-volume routine calls.

A practical rule: prototype on Sonnet 5, then route requests to Fable 5 at the point where Sonnet loses the plan or the context overflows. The switch is nothing more than changing the model field in the request body.

Quick Start: Your First Request in Python and curl

Onysoft exposes an OpenAI-compatible API, so you connect to Fable 5 with your existing OpenAI SDK by changing only the base_url and the key. Keys use the sk-ony- prefix.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.onysoft.com/v1",
    api_key="sk-ony-YOUR-KEY",
)

resp = client.chat.completions.create(
    model="anthropic/claude-fable-5",
    messages=[{"role": "user", "content": "Refactor this module and explain your plan step by step."}],
    stream=True,
)

for chunk in resp:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="")

The same request with curl:

curl https://api.onysoft.com/v1/chat/completions \
  -H "Authorization: Bearer sk-ony-YOUR-KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-fable-5",
    "messages": [{"role": "user", "content": "Hello Fable 5"}]
  }'

Because Fable 5 can reason for a long time on hard tasks, keep streaming on. See the API documentation for all parameters, or try it without writing any code in the Playground.

Cost Strategy: Reserve the Premium Model for Premium Work

Fable 5 is a flagship, and its per-token cost sits above Sonnet and Haiku. The most efficient setup is therefore not sending everything to Fable 5, but tiered model routing:

• Routine work (summaries, classification, simple extraction) → anthropic/claude-haiku-4.5 or openai/gpt-5.4-mini
• Everyday development and agent steps → anthropic/claude-sonnet-5
• Critical, long, context-heavy tasks → anthropic/claude-fable-5

The 1M context is tempting, but input tokens are billed too — fill the window because the task needs it, not because it fits. Current pricing is published in the model catalog, and you can model your own traffic with the cost calculator. There is no subscription on Onysoft: you pay as you go from a prepaid balance, so an experiment costs exactly what the experiment used.

Step-by-Step Access, From Turkey or Anywhere

You do not need a separate Anthropic account, a specific card, or a VPN to use Fable 5 — one Onysoft account covers it:

1. Sign up — an email address gets you an account in minutes.
2. Top up your balance — pay-as-you-go with no subscription. If you are in Turkey, you pay in TL at the central-bank rate with corporate e-invoicing.
3. Create your sk-ony- prefixed API key from the dashboard.
4. Pick anthropic/claude-fable-5 in the Playground and test it without writing code.
5. Point your app's base_url to https://api.onysoft.com/v1 and ship.

The same key unlocks far more than the Claude family: GPT-5.5, Gemini 3.1 Pro, DeepSeek v4 Pro, plus video, image, and music generation — 708+ models behind one OpenAI-compatible endpoint, operated by Izmir-based Onysoft Veri Merkezi A.S. with KVKK (Turkish data-protection) compliance and 24/7 support.