Compare

Opus 5 or Fable 5: which Anthropic model to take

For code, Opus 5, and for most other work, Opus 5 as well. Fable 5 charges twice as much while carrying the same context window, the same output ceiling and the same delivery routes; what the extra money buys is one place higher on the general text leaderboard, a gap of 12 points.

Specifications where both sides have a verified number

Claude Fable 5 Claude Opus 5
API model ID claude-fable-5 claude-opus-5
Context window 1 million tokens 1 million tokens
Max output 128k tokens 128k tokens
Input price $10 per million tokens $5 per million tokens
Output price $50 per million tokens $25 per million tokens
Knowledge cutoff January 2026 May 2026
Where it ships Claude API, Amazon Bedrock, Claude Platform on AWS, Google Cloud, Microsoft Foundry Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry

A row where one side has no verified number is left out entirely, because readers read a half-empty row in favour of the filled side.

The table above is the argument

In this pair both context windows are one million tokens, both output ceilings are 128,000 tokens, and both ship through the same routes. Two rows genuinely differ: price, $5 and $25 against $10 and $50, and knowledge cutoff, May 2026 against January 2026. The dearer model also holds the older knowledge.

What those 12 points are worth

On the Arena text leaderboard Fable 5 is first at 1506 and Opus 5 sits just under it at 1494. The same board prints a plus or minus 5 range beside both numbers, so the distance is not large. On the web-dev leaderboard the order reverses and the gap is more than five times wider: 1692 against 1628. If your work is code, the cheaper model is also the better one and the argument ends there.

Three things no table shows

Fable 5 is the only model marked slower in Anthropic own table. Its adaptive thinking is always on, so there is no lever for cutting reasoning cost. And the 300K batch output beta lists Opus 5 and four other models, but not Fable 5. All three are in the vendor documentation and none is on a sales page.

Pick this if Claude Fable 5

  • Your work is long general prose and you have measured a real difference on your own material.
  • Being slower does not bother you and neither does twice the price.
  • You need Claude Platform on AWS, which appears only in Fable 5 list of delivery routes.

Pick this if Claude Opus 5

  • Your work is code or multi-step agents. The gap there is 64 points in Opus 5 favour.
  • You want to halve the cost per task without giving up the window or the output ceiling.
  • You need the fresher knowledge: May 2026 against January 2026.

Questions people actually ask

Is Fable 5 smarter than Opus 5

It depends on the job, and both answers have a published number. On general text Fable 5 leads by twelve points; on web development Opus 5 leads by sixty-four. Smarter, without naming the task, means nothing.

Why does Fable 5 cost double

We do not know and we will not guess. Anthropic publishes no reason, and the published specifications of the two are identical on context window, output ceiling and delivery routes. The only measurable difference is where each one sits on the two leaderboards.