HunterAlphaHub
OpenRouter model reference Facts from the public catalogue, dated and labelled
Jev Laya Fable 5.2 Verified
2026-11-13

Model · Anthropic

Claude Fable 5.2

Claude Fable 5.2 is served by Anthropic through OpenRouter. It provides 1M tokens of context and supports text, vision, files inputs.

Snapshot 2026-11-13 anthropic/claude-fable-5.2 Edited by the Hunter Alpha Hub desk Fable 5.2, the effort dial and the demo wall →
Context window
1M tokens
Input / 1M tokens
$8.00
Output / 1M tokens
$40.00

Key facts

Provider Anthropic
OpenRouter ID anthropic/claude-fable-5.2
Modalities Text, Vision, Files
Best for Coding, Agents, Long Context
Source openrouter.ai/api/v1/models Verified 2026-11-13 Drift-checked daily 06:00 CST

Our read on Claude Fable 5.2

Fable 5.2 is the row you reach for when a task has already defeated the cheaper model, not the row you make the product default. It has the same 1M-token window and 128K output ceiling as Opus 5.5, but lists at $8 / $40 per million, exactly twice the price. Anthropic's own guidance starts most coding and knowledge work on Opus and moves upward only when the scope or difficulty justifies it. Take that advice seriously: paying twice as much for every routine request is a simple way to make a capable model look uneconomical.

The interesting part is the effort dial. High effort scored roughly level with Opus 5.5 at Max on Anthropic's FrontierCode 2.0 comparison for about half the cost per task; Xhigh is the setting behind the 49.1% headline. Artificial Analysis then ran the public model at Max and measured 46.8 minutes and $26.54 per intelligence-index task, with 267 seconds to the first answer token. That is a benchmark and long-batch profile, not a chat profile. Start with High, reserve Xhigh for work that survived cheaper attempts, and treat Max as a deliberate purchase.

Choose it when

A hard migration, a long unattended coding or research run, or an agent loop where one failed attempt costs more than the tokens. It is also sensible as an advisor or orchestrator that delegates routine steps to cheaper models.

Skip it when

Routine chat, high-volume extraction, well-scoped feature work, or anything a person is waiting on. Opus 5.5 has the same window at half the price; Sonnet 5.5 is cheaper again.

Closest call. Claude Opus 5.5 : Same 1M-token context and 128K output at $4 / $20 per million. Fable buys the harder reasoning tier, not more capacity, so the upgrade should be earned on your own task set.

Price, context, output limit, effort support and cache prices were read from OpenRouter on 2026-11-13. Anthropic's benchmark table and Artificial Analysis's independent Max-effort run are linked from the Fable topic. No private benchmark was run for this note.

Cost estimates

These examples use the 2026-11-13 input and output prices. Cache discounts, provider surcharges and tool fees are excluded.

Workload Input Output Estimated cost
Quick test 20,000 5,000 $0.360
Product chat month 500,000 125,000 $9.00
Document batch 5,000,000 1,000,000 $80.00

Strengths

  • ✓ Anthropic's frontier row for long-horizon agent work, released 2026-11-09 with a 1M-token window and a 128K output ceiling
  • ✓ Five reasoning-effort levels, including a 49.1% Xhigh result on Anthropic's FrontierCode 2.0 launch benchmark
  • ✓ Cached reads at $0.20 per million, which is where a long agent session with a stable prompt spends most of its input

Limitations

  • ! Twice Opus 5.5's token price at $8 / $40 per million, and Artificial Analysis's Max-effort run averaged $26.54 per task
  • ! Max effort is slow: 46.8 minutes per Artificial Analysis task and 267 seconds to the first answer token
  • ! General-access Fable carries cyber and biology safeguards; the less-restricted Mythos 5.2 is available only through Anthropic's verification programmes

OpenRouter API quickstart

Run these examples from your server or edge function. Keep the OpenRouter API key private and do not expose it in browser code.

cURL

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-fable-5.2",
    "messages": [{ "role": "user", "content": "Summarize this document in five bullet points." }],
    "max_tokens": 500
  }'

JavaScript / TypeScript

const response = await fetch("https://openrouter.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.OPENROUTER_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "anthropic/claude-fable-5.2",
    messages: [{ role: "user", content: "Summarize this document in five bullet points." }],
    max_tokens: 500,
  }),
});

const data = await response.json();
console.log(data.choices[0].message.content);

Compare models View on OpenRouter ↗

Claude Fable 5.2 FAQ

What is Claude Fable 5.2?

Claude Fable 5.2 is a Anthropic model available through OpenRouter. It supports text, vision, files inputs and is suited for coding, agents, long context workloads.

How much does Claude Fable 5.2 cost on OpenRouter?

The 2026-11-13 snapshot lists $8.00 per 1M input tokens and $40.00 per 1M output tokens. Verify current pricing on OpenRouter before production.

How large is Claude Fable 5.2's context window?

Claude Fable 5.2 offers 1M tokens of context in the current snapshot. Provider-specific limits and surcharges may apply.

Pricing and context data are based on a 2026-11-13 snapshot. Provider limits can change; always verify on OpenRouter before production.

Send us a link

A project built with Jev, a post, a video, a correction, a tip. We open the link, check it says what you said it says, and write the entry ourselves.

Required a link, and an email to reply to. Optional everything else.

Add context — all optional

One or two sentences about what it does, in your words. We write the entry ourselves.

Cost, latency, a benchmark — anything you measured. We attribute these to you.

We store what you type and email it to ourselves. No IP address, no user agent, no referrer — the same rule as the mailing list.

What happens to what you send →