What is Claude Fable 5?
Fable 5 was Anthropic's first Mythos-class model for general use, a tier above Opus. It reads text and images, writes text, and works across a 1M-token context with up to 128K tokens of output, with thinking always on. Safety classifiers hand some requests to Opus models: offensive security work to Opus 4.8 and dual-use biology to Opus 5.
Anthropic now lists it as a legacy model. It still runs on the Claude API, the major clouds and in the Claude apps, but Fable 5.1 replaced it at the same token price with cheaper cache reads, and Anthropic tells most people to start with the cheaper Opus 5.5.
Is Claude Fable 5 still available?
Yes, but Anthropic lists it as a legacy model: it still runs on the API, the major clouds and the Claude apps, and Fable 5.1 is now the current Fable model.
- API and clouds
- Requests to claude-fable-5 still work at $10 / $50 per million tokens on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS.
- Claude apps
- Paid plans can still pick Fable 5 in the model picker, on the same terms as Fable 5.1.
- Retirement
- Anthropic has committed to run it for at least a year after launch and gives at least 60 days' notice before retiring a public model.
- Successor
- Fable 5.1 at the same token price, while Anthropic points most other work to Opus 5.5.
Who is Claude Fable 5 best for?
Claude Fable 5 is best for teams whose code already calls claude-fable-5, and for long coding, finance and vision jobs they have not yet moved to Fable 5.1.
Code already built on Fable 5
Integrations that call claude-fable-5 keep working while you test Fable 5.1, which rejects forced tool use and ties thinking blocks to the conversation.
Codebase-wide coding jobs
Anthropic cites a migration across a 50-million-line codebase finished in a day, and early users said it cracked problems Opus 4.8 could not.
Finance and document analysis
Anthropic reports a top score on Hebbia's finance benchmark, with gains in reading documents, charts and tables.
Vision-heavy tasks
It can pull exact numbers out of detailed scientific figures and rebuild a web app's code from screenshots alone.
What's included with Claude Fable 5?
Always-on thinking with effort levels, server-side fallback for refused requests, agent tools, half-price batch jobs, every major cloud and the Claude apps.
- Always-on thinking
- Thinking cannot be turned off; an effort setting, high by default, controls how deep it goes.
- Fallback for refused requests
- The fallbacks parameter retries a declined request on another Claude model, and fallback credit refunds the prompt-cache cost of switching.
- Agent tools
- Supports the memory tool, code execution, programmatic tool calling, compaction and task budgets in beta.
- Batch API
- Half-price input and output, $5 / $25 per million tokens, for jobs that can wait.
- Every major cloud
- Available on the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS.
- In the Claude apps
- Paid plans can pick it on web, desktop and mobile, and in Cowork, Claude Code, Claude Design, Claude for Microsoft 365 and Claude Tag.
How much does Claude Fable 5 cost?
Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens on the Claude API; in the Claude apps it needs a paid plan, with Max (from $100 a month) covering it up to half your weekly limit and Pro ($20 a month) charging it as usage credits.
- Cache reads $1, cache writes $12.50 (5-minute) or $20 (1-hour) per million tokens
- Batch API halves it to $5 / $25
- Model ID claude-fable-5
- Fable 5 runs only on pay-as-you-go usage credits
- Usage credits are billed at standard API rates
- Opus, Sonnet and Haiku stay within the plan's limits
- Fable 5 and 5.1 included, up to 50% of your weekly usage limit
- 5x Pro usage for $100, or 20x for $200
- Usage credits keep Fable running past that cap
- Fable included up to 50% of each member's weekly limit
- Standard seats ($25) use Fable only through usage credits
- For teams of 2 to 150
The Free plan has no Fable access. Paying yearly brings Pro to $17 a month ($200 up front) and Team seats to $20 (standard) or $100 (premium) a month; Max is billed monthly only. Enterprise costs $20 per seat per month, billed yearly, plus usage at API rates, and usage-based Enterprise plans pay API rates for Fable. US-only inference costs 1.1x the input and output price. Anthropic now lists Fable 5 among its legacy models: Fable 5.1 has the same $10 / $50 token price, with cache reads at $0.25 instead of $1 per million tokens. Full pricing on claude.com. We checked these prices on .
What changed in Fable 5.1?
Mostly cost and long-running work: Fable 5.1 keeps the token price, cuts cache reads by 75%, and does better on long coding, research and document jobs.
- Cost
- Cache reads drop from $1 to $0.25 per million tokens, which Anthropic estimates cuts typical bills by about 25% and agent-heavy ones by up to about 45%.
- Capability
- Anthropic reports gains in long coding sessions, documents, spreadsheets and slides, web research, vision and computer use.
- Training data
- Its reliable knowledge runs about five months later than Fable 5's.
- Breaking changes
- Forced tool use returns an error, and editing earlier turns invalidates its thinking blocks, so check your code before you switch.
How do readers rate Claude Fable 5?
No one has rated Claude Fable 5 yet. Be the first.
Sign in to rate. One rating per account.
Claude Fable 5 pros and cons
Early users found it strong on hard coding and reverse engineering, but safety fallbacks, fast use of plan limits and a cheaper successor count against it.
Pros
- Early users said it cracked hard coding and reverse-engineering problems that Opus 4.8 and OpenAI's Codex had failed on, in one case after weeks of trying.
- Some early testers said it needed fewer tokens than Opus 4.8 for the same agent work, one of them about half, which offsets part of its higher price.
- Led Anthropic's launch table on SWE-Bench Pro at 80.3%, against 69.2% for Opus 4.8, 58.6% for GPT-5.5 and 54.2% for Gemini 3.1 Pro.
- Beat Opus 4.8 on Anthropic's vision and computer-use tests: 29.8% against 22.5% on GDP.pdf without tools, and 85.0% against 83.4% on OSWorld-Verified.
Cons
- Safety classifiers send offensive security work to Opus 4.8 and dual-use biology to Opus 5, and early users hit false positives on code reviews, security tests and physics questions; Anthropic later cut biology fallbacks by about 85%.
- It uses up plan limits fast: several Max subscribers said a single session emptied their five-hour window.
- Fable 5.1 costs the same per token with cache reads at $0.25 instead of $1, about 25% less on typical work by Anthropic's estimate, and Opus 5.5 costs $4 / $20.
- Pro and Team standard seats can use it only through paid usage credits, and it needs 30-day data retention, with zero data retention only when Anthropic approves it.
What are the best Claude Fable 5 alternatives?
Fable 5.1 for the same work at lower cost, Sonnet 5 at a fifth of the price, and GPT-6 Astra for a rival frontier model at the same token price.
Claude Fable 5.1PaidChoose Claude Fable 5.1 for any new work: it keeps the $10 / $50 price, cuts cache reads from $1 to $0.25 per million tokens, and has newer training data.
Claude Sonnet 5FreeChoose it when price matters more than peak quality: at $2 / $10 per million tokens it costs a fifth as much, with the same 1M context, and it is on every Claude plan, Free included.
GPT-6 AstraPaidChoose it when you want a rival frontier model at the same $10 / $50 API price, or some use inside the $20 ChatGPT Plus plan, where Claude Pro bills Fable as extra usage.
Questions about Claude Fable 5
What is Claude Fable 5?
Claude Fable 5 is Anthropic's first Mythos-class model for general use, built for long coding, research and document jobs; Anthropic now lists it as legacy, with Fable 5.1 as the current Fable model.
Who is Claude Fable 5 for?
Claude Fable 5 is best for teams whose code already calls claude-fable-5, and for long coding, finance and vision jobs they have not yet moved to Fable 5.1.
Is Claude Fable 5 free?
No. Claude Fable 5 is a paid tool, from $10 / $50 per million input / output tokens.
How much does Claude Fable 5 cost?
Claude Fable 5's paid plans are Claude API at $10 / $50 per million input / output tokens, Pro at $20 per month, Max at From $100 per month and Team premium seat at $125 per seat per month.
What are the best Claude Fable 5 alternatives?
The closest alternatives on Recaply are Claude Fable 5.1, Claude Sonnet 5 and GPT-6 Astra.
Where can I try Claude Fable 5?
On its official website, anthropic.com.
Is Claude Fable 5 included in Claude Pro?
No. On Claude Pro, Fable 5 runs only on usage credits, which are billed at standard API rates on top of the $20 monthly fee. Max plans and Team premium seats include Fable 5 up to half of the weekly usage limit, and the Free plan has no Fable access.
Why did Claude switch from Fable 5 to Opus in my conversation?
A safety classifier flagged the request, usually offensive security or dual-use biology, and Claude re-ran it on Opus 4.8 or Opus 5. The checks also read files, memory, connectors and search results, so text you did not type can trigger a switch.
Is Claude Fable 5 the same as Claude Mythos 5?
It is the same underlying model. Mythos 5 has the safety classifiers lifted in some areas and goes only to approved Project Glasswing partners, while Fable 5 is open to every paying customer at the same $10 / $50 price.
Does Claude Fable 5 work with zero data retention?
Not by default. Anthropic requires 30-day retention for Fable 5 traffic and allows zero data retention only when it expressly authorizes it. Anthropic says it will not use that retained data to train new Claude models.
Our verdict
Claude Fable 5 still works, but there is little reason to start anything new on it.
Fable 5.1 keeps the $10 / $50 token price, cuts cache reads by 75% and scores higher on Anthropic's tests, and Opus 5.5 handles most work at $4 / $20.
- Skip it for
- New projects, offensive security and professional biology work, where the safeguards hand requests to Opus models.
- Check first
- Before you move off it, test your code against Fable 5.1's changes to forced tool use and thinking blocks.
Claude Sonnet 4.6
Claude Sonnet 5.5
Claude Opus 4.7
Claude Opus 4.5