Anthropic has introduced Claude Fable 5.2 and Claude Mythos 5.2. Both models are based on the same base model. Fable 5.2 is open to all users and comes with safeguards for cybersecurity and biology that decline risky requests. Mythos 5.2 operates without some of these safeguards and is reserved for organizations that have passed Anthropic's vetting programs for cybersecurity or life sciences.
For everyday use, Anthropic continues to recommend Claude Opus 5.5. Fable and Mythos are intended for particularly demanding tasks, while simpler ones are handled by Sonnet 5.5 or Haiku 5.5. For most tasks in design, office work, and marketing, Opus thus remains the default.
What changes compared to Fable 5?
The list price in the API is 20 percent below the launch price of Fable 5 for both input tokens and output tokens. Anthropic prices in US dollars per million tokens; no euro price is available.
In terms of speed, Anthropic promises responses that come up to 30 percent faster than with Fable 5.1 and Mythos 5.1. According to Anthropic, an optional fast mode works up to 2.5 times as fast. The mode costs double the normal price. It runs in Claude Code, on the Claude Platform, in Claude on Google Cloud, and in Microsoft Foundry. Anthropic states the launch price of Fable 5 in the announcement from June 2026.
The safeguards kick in less often. Opus 4.8 answers blocked cybersecurity requests, and Opus 5 answers blocked biology requests. Compared to Fable 5.1, Anthropic counts around 55 percent fewer blocked requests. Tasks from design, office work, and marketing rarely touch on these topics. The refusals are therefore noticeable mainly in security research and biology.
Which model suits which task?
Anthropic specifies a clear order: Opus 5.5 as the daily standard, Fable 5.2 for particularly demanding tasks, Sonnet 5.5 or Haiku 5.5 for simpler ones.
Which Claude model for what?
Anthropic's recommendation at the release on November 9, 2026
What do you need the model for?
- Everyday use, programming, knowledge tasksClaude Opus 5.5
- Particularly demanding, open-ended tasksClaude Fable 5.2
- Simpler tasksSonnet 5.5 or Haiku 5.5
Mythos 5.2 is open only to vetted organizations
Source: Anthropic
The manufacturer's tests show where Fable 5.2 is ahead. In computer operation (OSWorld 2.1), the model reaches 89.3 percent, while Opus 5.5 gets 87.2 percent. In the Remote Labor Index, Fable 5.2 reaches 41.3 percent and Opus 5.5 21.3 percent. The gap is small in computer operation and large in the Remote Labor Index.
The list price says little about the bill. At the highest effort level, Fable 5.2 reaches 65 points in the Intelligence Index, according to Artificial Analysis. What a task costs depends on token consumption and the chosen effort level.
What this means for you
If you use Claude with a Pro subscription, Fable is included with up to 50 percent of the weekly usage limit, according to Anthropic. It's best to try Fable on a task on which Opus 5.5 recently failed. For everything else, the standard remains the better choice. In the Claude apps and in Claude Code, the default effort level is "Medium".
If you build automations via the API, calculate the cost per solved task and test the same task once with Opus 5.5 and once with Fable 5.2. The fast mode is only worthwhile if waiting time costs you more than double the price.
Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026.
Sources
