Anthropic just reshuffled its model lineup, and if you use Claude regularly, the change affects which model you should be reaching for — and how fast you’ll burn through your credits.
A New Tier at the Top
For a while, Claude’s lineup was easy to understand:
- Haiku for speed and low cost
- Sonnet for everyday productivity
- Opus for complex work
Claude Fable 5 adds a new tier above Opus.
Claude Fable 5 is built for demanding reasoning, coding, long-context analysis, and long-horizon agentic work. The official API model ID is:
claude-fable-5Anthropic also announced Claude Mythos 5, which shares the same capabilities but is only available through limited release. For most users, developers, and teams, Claude Fable 5 is the generally available option.
What Actually Changed?
The biggest change is not just better benchmark performance. It is the type of work Claude Fable 5 is built for.
The main upgrades are:
- stronger reasoning for difficult problems
- better support for long-running agentic tasks
- a 1M token context window
- up to 128k output tokens per request
- adaptive thinking always on
- effort controls for reasoning depth
- new refusal and fallback behavior
- higher pricing than Opus 4.8: Fable 5 costs about twice as much as Opus 4.8 at the API level
In plain English: Fable 5 is for work where quality matters more than speed or cost.
When to Use Claude Fable 5
Use Claude Fable 5 when the task is complex, long, or expensive to get wrong.
Good use cases include:
- advanced coding problems
- long-context document review
- research synthesis
- debugging subtle software issues
- multi-step planning
- technical architecture decisions
- agentic workflows
- final review of important work
Use a cheaper model when the task is simple.
For emails, outlines, quick summaries, brainstorming, and light rewriting, Sonnet or Haiku will usually be enough.
Reasoning Effort: The Dial to Know
Claude Fable 5 supports an effort setting that controls the trade-off between capability, latency, and token use.
The available levels are:
lowmediumhighxhighmax
Anthropic recommends starting with high, which is the default. Use xhigh for difficult coding, deep analysis, or long-horizon agentic work. Use medium or low when speed and cost matter more.
A simple rule:
High = strong default
Xhigh = hard work
Max = only for the toughest tasksSafety, Refusals, and Fallbacks
Claude Fable 5 includes safety classifiers that can decline certain requests.
In the API, a refusal can return as a successful response with:
stop_reason: "refusal"That means developers should handle refusals as part of the normal application flow, not only as errors.
Anthropic also supports fallback options, so a refused Fable 5 request can be retried on another Claude model. For regular productivity users, this usually will not matter much. For developers, it is important to design around it.
Data Retention Note
Anthropic says Claude Fable 5 and Claude Mythos 5 carry 30-day data retention and are not available under zero data retention.
For casual users, this may not change much. For businesses, developers, and teams handling sensitive information, it is worth reviewing before adopting Fable 5 in internal workflows.
Which Claude Model Should You Use?
Use Claude Fable 5 for the hardest reasoning, coding, long-context, and agentic tasks.
Use Claude Opus 4.8 for strong complex work at a lower cost than Fable 5.
Use Claude Sonnet 4.6 for everyday productivity, writing, coding help, and workflow automation.
Use Claude Haiku 4.5 for fast, low-cost tasks.
In short:
Fable 5 = hardest work
Opus = complex work
Sonnet = daily productivity
Haiku = fast and efficient tasksThe Practical Takeaway
Claude Fable 5 is powerful, but it should not be your default for everything.
Use Sonnet for drafts, summaries, brainstorming, and normal productivity. Use Opus when you need stronger reasoning. Use Fable 5 when the task is difficult enough that a better answer is worth the extra cost.
The best workflow is not “always use the strongest model.”
The best workflow is:
Use the cheapest model that can do the job well, and save Fable 5 for work that truly needs it.
FAQ
What is Claude Fable 5?
Claude Fable 5 is Anthropic’s most capable widely released Claude model, built for demanding reasoning, coding, long-context analysis, and long-horizon agentic work.
How much does Claude Fable 5 cost?
Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens at the API level.
Is Claude Fable 5 more expensive than Opus?
Yes. Claude Fable 5 costs about twice as much as Claude Opus 4.8 at the API level.
Should I use Claude Fable 5 every day?
Not for everything. Use it for complex work. For everyday writing, summarizing, and brainstorming, Sonnet is usually more cost-effective.
What effort setting should I use?
Start with high. Use xhigh for harder work and max only for the most difficult tasks.
