Put Claude to work on real tasks.
Claude Fable 5.1 is Anthropic’s current flagship model. It sits at the top of a line that runs Mythos, Fable, Opus, Sonnet and Haiku, and is positioned for long-running knowledge work and coding rather than for quick questions.
Pricing, benchmark figures, safeguards, retention policy and access routes on this page come from Anthropic’s own Claude Fable page. The two gallery figures are the ones Anthropic publishes there.
The two figures Anthropic publishes for Fable 5.1
Anthropic’s page for Claude Fable 5.1 carries two published graphics: the model’s key art and the benchmark table its performance claims rest on. Everything else on that page is a customer or partner logo, so there is no gallery of generated output to show — this section is what the vendor actually publishes.
Both figures belong to Anthropic and are reproduced with credit, from the Claude Fable page. Anthropic’s Claude Fable page
Fable 5.1 in Anthropic’s own words
Anthropic calls Fable 5.1 “our most capable generally available model” and places it in a Mythos tier above the rest of the line. The framing is work that runs for hours rather than turns: the page says it avoids easy-seeming shortcuts, fixes root causes rather than symptoms, and keeps you updated as it goes.
The top of a public line
Anthropic sells five model lines — Mythos, Fable, Opus, Sonnet and Haiku — spanning speed, price and capability. Fable 5.1 is the most capable one generally available; the Mythos tier above it is reached through Anthropic’s trusted-access programmes rather than self-service.
Built for hours, not turns
The page’s examples are jobs that span applications and run unattended: a backlog in Cowork, requests picked up in Slack through Claude Tag, a browser operated on your behalf, or a managed agent on the Claude Platform.
Safeguards that reroute rather than refuse
Cybersecurity and biology queries that trip the model’s safeguards are answered by less capable models instead — Opus 4.8 for cyber, Opus 5 for biology — and Anthropic says rerouted requests are not billed at Fable prices.
Retention by default
Using Fable requires 30-day data retention for safety monitoring. Enterprise customers eligible for Enterprise Frontier Safeguards can keep data in their own cloud infrastructure with human review done by the customer, and until that programme opens, eligible customers can run Fable 5.1 with zero data retention.
The four kinds of work Anthropic names
Anthropic groups Fable 5.1’s intended work under four headings and describes each one in prose rather than demonstrating it with output. That is why this page shows the model’s benchmark table instead of a gallery of samples: the vendor publishes charts, not generations.
Agents
Jobs that take hours and span several applications. The page describes the model planning the work, using the tools it needs, recovering when a step fails and reporting progress without being asked.
Coding
Features that run across an entire codebase, code review, performance work and multi-day autonomous sessions. Anthropic says it writes its own tests to check its work and uses vision to compare output against the original design.
Enterprise workflows
Complex, multi-stage knowledge work handled with minimal oversight, from research and analysis through to deliverables a team can review, rather than supervising every step.
Vision
Diagrams, charts and tables nested inside files and PDFs, for document-heavy work in finance, legal, analytics and architecture — the same vision capability the model uses to check its own coding output against a goal.
What Anthropic publishes about Fable 5.1
Only facts Anthropic states on its own pages appear here. Where the vendor gives no number — a context window, a throughput figure — this table does not invent one.
- Developer
- Anthropic
- Current model
- Claude Fable 5.1
- Tier
- Mythos-level, and the most capable generally available model
- API model id
- claude-fable-5-1
- Input price
- $10 per million tokens
- Output price
- $50 per million tokens
- Cache reads
- $0.25 per million tokens, which Anthropic puts at 75% less than Fable 5
- US-only inference
- 1.1x input and output pricing
- Consumer availability
- Pro, Max, Team and Enterprise plans
- Developer availability
- Claude Platform, available marketplaces, Amazon Web Services, Google Cloud, Microsoft Foundry
- Default data retention
- 30 days, for safety monitoring
- Benchmark conditions
- Evaluated with production safeguards enabled
Documented access routes
Four ways in, all of them documented on the Fable page itself: the consumer plans, the first-party API, the cloud marketplaces, and the system card for anyone who needs the safety detail before committing.
Claude Fable page
The vendor page this whole article is drawn from: pricing, safeguards, benchmarks and the two gallery figures.
OpenClaude
The consumer app, where Pro, Max, Team and Enterprise plans reach Fable 5.1.
OpenClaude Platform
The developer route, with the API model id claude-fable-5-1.
OpenClaude pricing
Anthropic’s plan and token pricing, including the tiers Fable 5.1 arrives on.
OpenFable 5.1 system card
The safety results Anthropic publishes alongside the model, referenced from the benchmarks section.
OpenClaude Fable 5.1 questions
What is Claude Fable 5.1?
Anthropic’s current flagship and its most capable generally available model. The page describes it as a Mythos-level model built for ambitious, long-running work, and it sits above Fable 5, Opus 5 and Sonnet in the line-up.
How much does Claude Fable 5.1 cost?
$10 per million input tokens and $50 per million output tokens. Cache reads cost $0.25 per million tokens, which Anthropic says is 75% less than Fable 5 and cuts typical workloads by an estimated 25% — and heavily agentic ones by up to about 45%. Workloads pinned to US-only inference pay 1.1x on both.
Is Fable 5.1 the most capable Claude model?
It is the most capable generally available one. Anthropic keeps a Mythos tier above it for cybersecurity and biology research, reachable through trusted-access programmes rather than a public plan.
What happens when a query hits the safeguards?
It is rerouted rather than answered by Fable: cybersecurity queries to Opus 4.8 and biology queries to Opus 5. Anthropic says rerouted requests are not charged at Fable prices, and that the rerouting lowers Fable 5.1’s scores on benchmarks that include those domains.
How long does Anthropic keep Fable 5.1 data?
Thirty days by default, for safety monitoring. Enterprise customers eligible for Enterprise Frontier Safeguards can store data on their own cloud infrastructure with human review done by the customer; until that programme is available, eligible customers can use Fable 5.1 with zero data retention.
Where can I use Claude Fable 5.1?
On Pro, Max, Team and Enterprise plans, and for developers through the Claude Platform, available marketplaces, Amazon Web Services, Google Cloud and Microsoft Foundry. The API model id is claude-fable-5-1.
Latest Claude Fable 5.1 articles
Browse every published Claude Fable 5.1 article, from introductions to practical guides and developer documentation.
-
Claude Fable 5.1: What Changed and How to Move
Anthropic's new model costs the same per token as Fable 5 but reads cached prompts at a quarter of the price. Specs, the real change, and two prompts to copy.
Read article -
Claude Prompt Engineering: What the 2026 Best Practices Change
The Claude prompting techniques Anthropic recommends in 2026, the older advice it now plays down, and runnable code for prefill and chaining.
Read article -
How to Use Claude for Free: What the $0 Plan Covers
The free tier is $0. Which models and features it includes, how the five-hour session limit works, and three prompts a beginner can copy.
Read article -
Claude vs DeepSeek: Two Rate Cards, Read Side by Side
A Claude vs DeepSeek comparison built only from vendor documentation: both rate cards side by side, plus what each side leaves unpublished.
Read article -
How to Use Claude Code in the Terminal: The First Ten Minutes
Install Claude Code from the official one-liners, start a session, run a single task without one, and fix the PATH error that stops most first runs.
Read article -
Claude vs ChatGPT: Compare Them Without Borrowed Numbers
A Claude vs ChatGPT comparison built only from vendor pages: the Claude column filled from Anthropic, the ChatGPT column left empty, plus the audit prompt.
Read article -
Claude vs Gemini: Comparing Two Assistants Honestly
A Claude vs Gemini comparison built from vendor pages only: the Claude column filled from Anthropic, the Gemini column left empty, and the audit prompt.
Read article -
Is Claude AI Free? What the $0 Plan Actually Gives You
Anthropic's pricing page lists a Free tier at $0. What it includes, which models it gates, how the five-hour windows work, and what is not published.
Read article -
Claude Pricing Plans: Pro vs Max 5x vs Max 20x
The official Claude price table, the tier rules that matter more than price, and how to choose between Pro, Max 5x and Max 20x using published figures only.
Read article -
How to Use Claude AI
Where Claude runs, how a first conversation goes, one worked example with the whole prompt, and the current model line with Anthropic prices.
Read article -
Claude API: Getting Started
Call the Claude API with a working curl request and a Python SDK example, plus the Fable 5.1 model id, published pricing and what Anthropic has not published.
Read article