Claude Fable 5.2: What's New, Business Workflows, and Alternatives (2026)

Claude Fable 5.2 is Anthropic's newest Fable model, released on November 9, 2026. It runs faster than Fable 5.1 and scores much higher on Anthropic's business workflow benchmark. Its API price is $8 and $40 per million tokens. Anthropic still recommends Claude Opus 5.5 as the starting point for most everyday work.
This guide covers what Anthropic announced, what changed for business tasks, who can use it, and what it costs. It ends with alternatives, including a way to turn data into finished reports without picking a model. Every detail comes from vendor pages read on November 10, 2026.
What Anthropic announced on November 9
Anthropic released two models on the same day. Claude Fable 5.2 is available to all users. Claude Mythos 5.2 is for approved organizations in cybersecurity and biomedical research.
They are the same model underneath. Anthropic's post says "Claude Fable and Claude Mythos are the same underlying model." Fable has stronger safeguards against misuse, which is why Anthropic can offer it widely.
The headline claims are short:
- Better results. Improvements "in areas like coding, reasoning, and vision."
- Faster output. Outputs arrive "up to 30% faster" than the 5.1 models.
- API price. $8 per million input tokens and $40 per million output tokens.
- Fast mode. An optional mode "speeds up token generation by up to 2.5 times."
- Fewer refusals. About 55% fewer blocks than Fable 5.1 in Anthropic's early testing.
Anthropic also described prerelease tests. The model helped engineers finish software migrations and design three working electronic devices. It also "formalized the proof of Fermat's Last Theorem in 30 hours."
What changed for business workflows
Most coverage focuses on coding and science. For office work, one row in Anthropic's table matters most. It is AutomationBench, which Anthropic labels "Business workflows."
| Model | AutomationBench score |
|---|---|
| Claude Fable 5.2 | 51.4% |
| Claude Fable 5.1 | 31.4% |
| Claude Opus 5.5 | 42.5% |
| GPT-6 Astra | 41.4% |
| Gemini 4 Carbon | 53.9% |
That is a 20-point jump over Fable 5.1. It also puts the new Fable ahead of Opus 5.5 and GPT-6 Astra on this test.
It is not the top score, though. Gemini 4 Carbon scored 53.9% in Anthropic's own table. Fable 5.2 leads on most other rows, but business workflows is the exception.
Benchmarks are a starting point, not a verdict. A score tells you how a model did on a set of tasks. It does not tell you how it will handle your own files, templates, and approval steps.
The other benchmark rows
Anthropic says the new models "are a step above Opus 5.5 in almost all the capabilities we evaluated." Here are some rows from its table.
| Test | Fable 5.2 | Fable 5.1 | Opus 5.5 |
|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% | 32.4% | 45.5% |
| FrontierSWE v2 (agentic coding) | 69.9% | 56.3% | 62.3% |
| Terminal-Bench-Science 0.1 (scientific research) | 74.6% | 52.6% | 58.7% |
The coding gains are large. On FrontierCode 2.0, Anthropic lists the Fable scores at its "Xhigh" effort setting. That matters, since higher effort settings take longer and use more tokens.
When to use it, and when not to
Anthropic is direct about this. Fable and Mythos "are best for the most ambitious, open-ended technical work."
For daily tasks, the advice is different. Anthropic writes that "For most coding and knowledge tasks, we recommend that users start with Claude Opus 5.5." Users can move up to Fable for harder work, or down to Sonnet 5.5 and Haiku 5.5 for simpler work.
The new Fable can also work with smaller models. Anthropic describes two setups:
- Advisor. Haiku, Sonnet, or Opus models "consult it to help solve harder problems."
- Orchestrator. The new Fable "delegates simpler tasks to smaller models."
For business users, this suggests a simple rule. Use Opus for routine reports and emails. Try Fable on long, multi-step jobs where a smarter plan saves rework.
Good candidates share a few traits. They span several files, follow many rules, and fail in costly ways when a step is missed. Examples include a month-end roll-up from several exports, a contract review against a checklist, or a migration of old reports to a new template.
Weak candidates are short and well defined. A one-paragraph email, a quick summary of a single file, or a formula fix rarely needs the strongest model. Paying more per token for those jobs adds cost without much gain.
A fair test keeps everything else the same. Use the same files, the same prompt, and the same reviewer for both models. Then compare three things: errors found in review, minutes spent fixing, and total cost.
A business workflow example
Here is how a team might test the model on a real workflow. The company and numbers are made up for illustration.
Northfield Dental runs twelve clinics. Each month, an operations analyst pulls bookings, cancellations, and revenue from three exports. She builds a summary for the regional managers. It takes about six hours.
Week 1. She gives Opus 5.5 the three exports and her last report. She asks it to repeat the same analysis for October. The draft is close, but two clinic names are mismatched across files.
Week 2. She runs the same prompt with Fable 5.2. This time the model flags the mismatched names before building the summary. The draft needs fewer fixes.
Week 3. She compares cost and time. Fable costs more per token, so she keeps it for the monthly roll-up only. Weekly updates stay on Opus.
Week 4. She writes down what changed. The roll-up now takes about four hours instead of six. Most of the saved time came from fewer name fixes, not from faster writing. She keeps the same three exports and the same prompt, so next month's test is easy to repeat.
The point is not that one model always wins. Test the step that breaks most often, and use the stronger model only there.
Safety, refusals, and data retention
Business users often hit a different problem with frontier models. A harmless request gets blocked because it looks like a risky one. Anthropic says this release improves on that.
The post says the new safeguards "are as robust as those of Fable 5.1 but produce fewer unnecessary blocks." In early testing across claude.ai, Claude Code, and Cowork, Anthropic saw "about 55% fewer blocks" than with Fable 5.1. The drop was about 60% in cybersecurity and 40% in biology.
When a request is still blocked, the user is not left with nothing. Anthropic says users "will instead receive a response from Opus 4.8 for cybersecurity and Opus 5 for biology."
Data retention is the other point to check. Anthropic says powerful models like this one use safeguards that look at patterns across many requests. For most customers, that means "30-day retention for safety monitoring."
Some enterprise customers have another path. Enterprise Frontier Safeguards runs safety monitoring on data stored in the customer's own infrastructure. Anthropic says this makes it "compatible with zero data retention." The program is in private beta, and enterprise customers request access through their account team.
If your team handles client or patient data, read these terms before rollout. The model choice affects where records sit, not only how good the answers are.
Anthropic also notes a softer change. Testers found the new models "communicate more clearly than past Fable and Mythos models." For reports that go to managers, clearer writing can save an editing pass.
Who can use it
Anthropic's post says the model "is available today on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure." Developers can call claude-fable-5-2 on the Claude Platform.
In the Claude apps, access depends on the plan. The Claude pricing page lists Fable access this way:
| Claude plan | Fable access, per the pricing page |
|---|---|
| Free | No |
| Pro | Paid with usage credits |
| Max | 50% of weekly limits |
| Ultra | 50% of weekly limits |
| Team | 50% of weekly limits on premium seats |
Mythos 5.2 is separate. Organizations apply through Anthropic's Cyber Verification Program or Life Sciences program.
What it costs
Prices below come from Anthropic's announcement and the Claude pricing page on November 10, 2026.
| Option | Price |
|---|---|
| Fable 5.2 API | $8 per million input tokens, $40 per million output tokens |
| Fable 5.2 fast mode | $16 per million input tokens, $80 per million output tokens |
| Fable 5.1 API (legacy) | $10 per million input tokens, $50 per million output tokens |
| Opus 5.5 API | $4 per million input tokens, $20 per million output tokens |
| Claude Pro | $20 billed monthly |
| Claude Max | $100 per month |
| Claude Ultra | $200 per month |
The new model costs twice as much per token as Opus 5.5. That is why Anthropic's advice to start with Opus matters for budgets.
Current limitations
The launch has clear edges, based on Anthropic's own pages:
- Not on Free. The Free Claude plan does not include Fable.
- Pro pays per use. On Pro, Fable is paid with usage credits rather than included.
- Not the business-workflow leader. Gemini 4 Carbon scored higher on AutomationBench in Anthropic's table.
- Higher effort costs more. Some top scores use the Xhigh effort setting.
- Built for hard technical work. Anthropic recommends Opus 5.5 for most daily tasks.
There is also a format gap. A better model writes a better draft. It still needs your data, your template, and a place where the final report or deck gets built.
Our earlier guide to Claude Fable 5.1 covers how that release handled decks.
Alternatives to Claude Fable 5.2
Other options cover parts of the same work. Prices come from each vendor's page on November 10, 2026.
| Option | Strength, per the vendor | Price |
|---|---|---|
| Claude Opus 5.5 | Anthropic's recommended "daily driver" | $4 / $20 per million tokens |
| GPT-6 Astra | 41.4% on AutomationBench in Anthropic's table | $10 / $50 per million tokens |
| Gemini 4 Carbon | 53.9% on AutomationBench in Anthropic's table | Not listed on Google's Gemini API pricing page on November 10 |
| Powerdrill Bloom | Reports, slides, and sheets from your data | Free plan; Pro $13.27/mo |
Claude Opus 5.5. It costs half as much per token. For most reports and emails, Anthropic itself suggests starting here.
GPT-6 Astra. OpenAI's API pricing page lists it at $10 per million input tokens and $50 per million output tokens. It scored close to Opus 5.5 on business workflows in Anthropic's table.
Gemini 4 Carbon. It had the top AutomationBench score in the table. Google's Gemini API pricing page did not list Carbon when we checked, so confirm access with Google.
Where Powerdrill Bloom fits
A model choice is only one part of a business workflow. The other part is turning data into a document people can use. Powerdrill Bloom handles that step without asking you to pick a model.
The AI Report Generator says "Simply upload your data file, and let the AI do the rest." The Business Review Generator builds "quarterly or monthly business review decks" from your data. The Excel to PPT page turns Excel, CSV, or PDF files into slides.
The pricing page lists a Free plan at $0.00 and Pro at $13.27/mo.
Try Powerdrill Bloom with last month's exports and the report you usually build by hand.
Frequently asked questions
When was Claude Fable 5.2 released?
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. Fable 5.2 is available to all users, and Mythos 5.2 is for approved organizations.
How much does Claude Fable 5.2 cost?
The API price is $8 per million input tokens and $40 per million output tokens. Fast mode costs $16 and $80. In the Claude apps, access depends on your plan.
Is Claude Fable 5.2 better than Opus 5.5?
On most benchmarks in Anthropic's table, yes. On AutomationBench, it scored 51.4% to Opus 5.5's 42.5%. Anthropic still recommends Opus 5.5 for most daily coding and knowledge tasks.
Is Claude Fable 5.2 on the Free plan?
No. The Claude pricing page lists no Fable access on Free. Pro pays for Fable with usage credits, and Max and Ultra include it for 50% of weekly limits.
What is the difference between Fable 5.2 and Mythos 5.2?
They are the same underlying model. Fable has stronger cybersecurity and biology safeguards, so it is open to everyone. Mythos is available to vetted organizations through Anthropic's verification programs.
Sources: Anthropic, Claude Fable 5.2 and Mythos 5.2 · Claude pricing · OpenAI API pricing · Gemini API pricing. Details read on November 10, 2026.