Services / Anthropic / Models
Claude models

Four models, and which to use when.

Claude is a family of models, not one. Anthropic's current line-up runs from Fable 5.1, built for multi-day autonomous work, through Opus 5.5 and Sonnet 5.5 to Haiku 5.5, its fastest and cheapest. The same models power every Claude surface, so choosing one is about the job: how hard it is, how long it runs and how fast and cheap it needs to be.

Compare the four →Controls that change cost

Model versions change often. These are the models Anthropic lists as current in November 2026.

Illustration, simplified. Positions are indicative, not measured.
How to choose

Match the model to the job.

Anthropic's own advice runs through its model pages: use the most capable model where judgement matters, and faster ones for volume, often in the same system.

01Hard, long, open-ended
Multi-day autonomous projects, ambitious codebase-wide work and frontier research: Fable 5.1.
02Serious daily work
Agentic coding, long-running agents and enterprise workflows that need precise instruction-following: Opus 5.5.
03Well-scoped work at scale
Everyday coding, customer-facing agents and multi-tool workflows that run repeatedly: Sonnet 5.5.
04Speed and volume
Real-time chat and voice, classification, summarisation, routing and sub-agents: Haiku 5.5.
05Several at once
A more intelligent model plans and hands subtasks to Haiku, so many agents run in parallel at lower cost.
06Effort, per task
Effort controls trade cost against intelligence task by task; Haiku 5.5 is the first Haiku to have them.
The family

The four models, as Anthropic describes them.

Each is available on the Claude Platform and on Amazon Web Services, Google Cloud and Microsoft Foundry.

ModelWhat it is for
Fable 5.1A Mythos-level model for the most ambitious, long-running projects: jobs that take hours across many applications, codebase-wide features, code review, performance work and multi-day autonomous sessions. It writes its own tests and uses vision to check outputs against the goal. Available in Claude on Pro, Max, Ultra, Team and Enterprise.
Opus 5.5A hybrid reasoning model with a 1M-token context window, built as the enterprise daily driver: long-running coding in large codebases, agents that coordinate sub-agents and use memory, financial analysis, and vision and computer use. Anthropic says it communicates clearly and follows writing rules. Available in Claude on Pro, Max, Ultra, Team and Enterprise.
Sonnet 5.5A hybrid reasoning model with a 1M-token context window: a faster, lower-cost complement to Opus for well-scoped coding, agent tasks such as investigation, review and drafting, design and visual work, and office tasks. Anyone can use it in Claude.
Haiku 5.5Anthropic's fastest and most efficient model, for high-volume, cost-sensitive work: classification, summarisation, real-time chat and voice, sub-agents, simple coding and repetitive browser and desktop automation. Selectable in Claude on every plan, and in Claude Code.

Claude Platform models and pricing

Mythos and Fable

Why there is a model most people cannot use.

Anthropic's most capable model class, Claude Mythos, is strong enough in cybersecurity and biology that Anthropic limits who can use it. Fable is how those capabilities are released more widely.

01Claude Mythos 5.1
Anthropic's most capable model class for cybersecurity and biology research, available only to vetted cyber-defenders and life scientists through trusted access programmes.
02Claude Fable 5.1
The same underlying model as Mythos 5.1, with safeguards for cybersecurity and biology. Flagged requests in those areas are routed to less capable models, and you are not charged Fable prices for them.
03Verification programmes
Security teams can apply to the Cyber Verification Program for advanced cyber capabilities on Opus 5.5, Sonnet 5.5 and Mythos 5.1; life-science researchers to the Life Sciences Verification Program.
04Retention for safety monitoring
Using Fable or Mythos requires 30-day data retention for safety monitoring by default. For Fable, enterprises eligible for Enterprise Frontier Safeguards can keep data on their own cloud instead.

Claude Security, Anthropic's code-scanning product for Claude Enterprise, now runs on Mythos 5.1.

Controls that change cost

The same model, cheaper or faster.

From Anthropic's model and platform pages. Prices change, so none are quoted here.

ControlWhat it does
EffortChoose how hard Claude works on a task. Anthropic's model pages describe fine-grained control over thinking effort for API users.
Prompt cachingReuse background knowledge and examples across requests at a much lower price for the cached part; Anthropic quotes cache reads separately on each model.
Batch processingSend large volumes asynchronously for a lower price when results are not needed in real time.
Fast modeFaster responses from Opus 5.5 at a higher price per token, in Claude Code and on the Claude Platform.
US-only inferenceFor workloads that must run in the United States, at a premium on input and output tokens.
Context windowUp to 1M tokens depending on the model. Longer prompts cost more; Haiku 5.5 is priced differently above 100K tokens.

© 2026 Mawkish Technologies. This guide, its diagrams and illustrations are protected by copyright. You’re welcome to learn from it and to quote it with credit and a link. Reproducing it, by hand or through AI, and publishing it as your own is not allowed. Read our copyright notice

Questions

Straight answers.

Which model should we use?

Start from the job. Anthropic positions Opus 5.5 as the enterprise daily driver, Sonnet 5.5 for well-scoped work at scale, Haiku 5.5 for speed and volume, and Fable 5.1 for the hardest long-running work. Many systems use more than one.

Are the same models in the Claude app and the API?

Yes. Anthropic says the same frontier models power every surface, from business apps to developer tools. Which models are available in the app depends on the plan.

What is the difference between Fable and Mythos?

They are the same underlying model. Mythos 5.1 is limited to vetted organisations because of its cybersecurity and biology capabilities; Fable 5.1 adds safeguards in those areas so it can be offered widely.

Do some models keep our data longer?

Yes. Anthropic says using Fable or Mythos requires 30-day retention for safety monitoring by default. For Fable, enterprises eligible for Enterprise Frontier Safeguards can keep data on their own cloud instead.

Why are there no prices or benchmark scores here?

They change with each release. Anthropic publishes current prices on its Claude Platform page and benchmarks on each model page, both linked below.

Checked against Anthropic's own pages in November 2026.

Across Claude

Go to another Claude page.

Claude Enterprise, the Claude Platform, Claude and Claude Code each have their own page. Visit them in any order. Choose one to go there.

You are currently viewing Claude models, a supporting page

Also on Claude

How it all fits together

Talk to us

Talk it through with us.

Tell us what you want Claude to do for your organisation, and we will help you work out where to start, safely.

  • Your goals, data and the systems involved
  • Claude Enterprise, the Platform, or both
  • A sensible first step

Anthropic, Claude, Claude Code and the other product names on these pages are trademarks of Anthropic, PBC. No Anthropic logo or artwork is used, and nothing here implies Anthropic's sponsorship or endorsement. Other names are the property of their owners. Screens shown are our own illustrations with sample data, not Anthropic screenshots. Product descriptions follow Anthropic's own published material and may change; check claude.com and anthropic.com for current details.

The Yorcombe Relay is our shorthand for looking after the details nobody sees in Claude models.

© 2026 Mawkish Technologies. This guide, its diagrams and illustrations are protected by copyright. You’re welcome to learn from it and to quote it with credit and a link. Reproducing it, by hand or through AI, and publishing it as your own is not allowed. Read our copyright notice