Claude Fable 5.2: What's New, What It Costs and Whether a Small Company Should Care
Claude Fable 5.2 launched on 9 November 2026: 20% cheaper at $8 / $40, a week included in Claude Pro, and why Opus 5.5 is still the right default for most small companies.
Žiga Flis · · 8 min read
Claude Fable 5.2 is Anthropic's top model for the hardest work. It launched on 9 November 2026 alongside Claude Mythos 5.2 and costs $8 per million input tokens and $40 per million output tokens on the API, 20% less than Fable 5.1. Claude Pro users can try it inside their plan until 16 November. My verdict: Opus 5.5 is still the right starting point for most small companies, and Fable 5.2 is now a cheaper tool for the handful of tasks where Opus falls short.
If you are wondering whether the news means you need to act: a model release is no reason to rebuild your processes. Below is what changed, who gets it on which plan, and what is worth testing this week. I checked every price and condition against Anthropic's own pages on 9 November 2026.
What's new in Fable 5.2
Fable 5.2 is the third Fable model in five months, after Fable 5 (9 June 2026) and Fable 5.1 (1 September 2026). Anthropic describes improvements "in coding, reasoning, and vision", and says testers found the models "communicate more clearly than past Fable and Mythos models". Anthropic also reports outputs up to 30% faster than Fable 5.1.
The specs stay familiar: a 1M-token context window, up to 128K output tokens, text and images in, and a knowledge cutoff of June 2026.
Anthropic calls Fable and Mythos "the same underlying model", with the same specs and pricing. Fable 5.2 is generally available, with safeguards for cybersecurity and biology. Mythos 5.2 is only available to vetted organisations through the Cyber Verification Program and the Life Sciences Verification Program. For a small company, Fable is the one that matters.
The announcement cites customer examples. At Box, while reviewing loan application files, Fable 5.2 flagged that salary deposits and a bank statement's balance "didn't match". At Oracle, its SQL returned the expected results, while two other models' SQL "incorrectly summed country-level summary amounts". The vendor picked these examples, but they show where a top model earns its price: checking whether figures across several documents actually agree.
Fable 5.1 and Fable 5 have not been retired. Anthropic lists them as legacy models that are still available.
What Fable 5.2 costs
The headline for your budget: the top model got cheaper.
| Model | Input (per 1M tokens) | Output (per 1M tokens) | Best for |
|---|---|---|---|
| Claude Fable 5.2 | $8 | $40 | demanding reasoning and long-horizon agentic work |
| Claude Fable 5.1 (legacy) | $10 | $50 | predecessor, still available |
| Claude Opus 5.5 | $4 | $20 | recommended default for most work |
| Claude Sonnet 5.5 | $2 | $10 | everyday work |
| Claude Haiku 5.5 | from $0.10 | from $0.50 | high-volume simple tasks |
Fable 5.2 now costs twice as much as Opus 5.5; with Fable 5.1 the gap was 2.5x. It is also cheaper than OpenAI's GPT-6 Astra ($10 / $50).
Cache reads cost $0.20 per million tokens, 5-minute cache writes $10 and 1-hour cache writes $16, and the Batch API is 50% off. Fast mode generates tokens up to 2.5x faster, but at double the price, $16 / $80. In office work speed is rarely the bottleneck, so for a small company that seldom pays off.
These are developer prices. For subscription costs, see Claude Pro vs Max vs Ultra price in 2026.
Who gets Fable 5.2 on which plan
The rules for Fable models did not change with 5.2:
- Free: no Fable models.
- Pro and Team standard seats: Fable runs on pay-as-you-go usage credits.
- Max ($100), Ultra ($200) and Team premium seats: up to 50% of your weekly usage limits can go to Fable at no extra cost. After that, usage credits or another model.
- Enterprise: Fable 5.2 stays off until an admin enables it. Seven days after launch, Anthropic adds it to roles that already have a Fable model enabled, unless the admin opts out.
Fable models use up limits faster than other models. The default model does not change, so you have to pick Fable in the model picker yourself. Claude Code needs version 2.1.321 or later.
Launch offer: Pro until 16 November
Pro and Team standard seats get all Fable models included in the plan until 16 November 2026 at 11:59 pm PT (8:59 am CET on 17 November). You can spend up to 50% of your weekly usage limits on Fable and you get one Fable limit reset. After that, Fable goes back to usage credits.
Max, Ultra and Team premium seats get one to three Fable limit resets depending on how long you have subscribed (one under six months, three over twelve months). They expire on 8 March 2027.
Paid billing had to be in place before 9 November 2026 at 10:00 am PT. A reset restores the weekly usage spent on Fable models; you use it under Settings > Usage > "Limit resets" > "Reset for free".
What Anthropic's numbers say
The results below are Anthropic's own benchmarks from the launch announcement, not independent measurements.
| Benchmark | Fable 5.2 | Fable 5.1 | Opus 5.5 | GPT-6 Astra |
|---|---|---|---|---|
| FrontierCode 2.0 (agentic coding) | 49.1% | 32.4% | 45.5% | 42.8% |
| FrontierSWE v2 (agentic coding) | 69.9% | 56.3% | 62.3% | 65.5% |
| AutomationBench (business workflows) | 51.4% | 31.4% | 42.5% | 41.4% |
| OSWorld 2.1 (computer use) | 89.3% | 83.5% | 87.2% | 79.5% |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% |
| Humanity's Last Exam (with tools) | 74.2% | 65.6% | 67.7% | 57.2% |
Three honest notes. The FrontierCode 2.0 score was reached at xhigh effort. On AutomationBench, the most relevant test for businesses, Fable 5.2 is not first: the same announcement puts Google's Gemini 4 Carbon at 53.9%. And on five of the six benchmarks the lead over Opus 5.5 is between two and nine percentage points, while the price is double.
Anthropic adds that on FrontierCode 2.0, Fable 5.2 at high effort scored 45.3%, close to Opus 5.5's best of 45.5%, "for roughly half the cost per task". So a pricier token does not automatically mean a pricier task. But it is one coding benchmark and a vendor claim, so test it on your own work.
Fable 5.2 or Opus 5.5 for a small company
Anthropic's models overview says: "If you're unsure which model to use, start with Claude Opus 5.5 for most workloads. Use Claude Fable 5.2 for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short." I agree, and it is what I tell teams in workshops.
In practice I use three rules:
- Opus 5.5 by default. Quotes, email, contract summaries, customer support and most analysis. For high-volume simple work, Sonnet 5.5 or Haiku 5.5 is enough.
- Fable 5.2 when Opus fails. If Opus at higher effort still does not get the result right, that is your signal for Fable. Typically these are long multi-step tasks, or checking data across several documents.
- On a subscription, count limits, not dollars. On Max and Ultra, Fable costs nothing extra up to 50% of your weekly limits, but it burns through them faster. Use it for the hard task, then switch back for routine work.
Limits worth knowing
Data retention. Fable 5.2 requires 30-day data retention and, in Anthropic's words, "isn't available under zero data retention unless expressly authorized by Anthropic". Anthropic names Enterprise Frontier Safeguards, still in private beta, as the planned route to compatibility.
Safeguards and model switching. Classifiers block high-risk cyber activity and most research biology tasks. In the Claude apps, a blocked request is answered by an older model instead: cyber by Opus 4.8, biology by Opus 5. From early testing, Anthropic reports "about 55% fewer blocks with Fable 5.2 compared with Fable 5.1".
For your own integration. The model ID is claude-fable-5-2, with retirement not sooner than 9 November 2027. Adaptive thinking is always on and cannot be turned off; the default effort is high. Forced tool use (tool_choice set to any or tool) returns a 400 error, and you should omit temperature, top_p and top_k. Anthropic also describes using Fable 5.2 as an advisor that smaller models consult, or as an orchestrator that hands simpler tasks to them.
What to do this week
- On Pro, test one hard, real task before 16 November. Take a task where Opus 5.5 disappointed you and run the same input through Fable 5.2.
- On Max or Ultra, check your resets under Settings > Usage. They last until March, so there is no rush.
- If you use Fable 5.1 on the API, test the switch on your own test cases. Fable 5.2 is cheaper.
- Don't rebuild your processes. Nothing will stop working. An automation that saves you three hours a week today will be just as good tomorrow.
Ready to move from news to actual use? In the Claude AI workshop for teams, your team does not learn which model tops a leaderboard. It builds something useful with the one it already has. Book a free 30-minute intro call, no commitment.
Related reading:
- Claude Pro vs Max vs Ultra price in 2026: which plan fits how you work.
- Claude AI for European SMBs in 2026: a practitioner's rollout guide.
- New AI models 2026: an overview for companies (in Slovenian): every current model and price in one table.
Sources
- Anthropic: Claude Fable and Mythos 5.2: release, improvements, price, fast mode, benchmarks, customer examples and safeguards; checked 9 November 2026.
- Claude Docs: Claude Fable 5.2: specs, cache and Batch API pricing, data retention, Mythos 5.2; checked 9 November 2026.
- Claude Docs: Models overview: model ID, retirement date, legacy models and model choice guidance; checked 9 November 2026.
- Claude Fable models on your plan (Anthropic Help Center): Fable by plan, Enterprise and Claude Code; checked 9 November 2026.
- Claude Fable 5.2 launch offers (Anthropic Help Center): launch offer, limit resets and eligibility; checked 9 November 2026.
Frequently asked questions
What is Claude Fable 5.2?
Claude Fable 5.2 is Anthropic's top model for demanding reasoning and long-horizon agentic work, released on 9 November 2026 as the successor to Fable 5.1. Anthropic reports improvements in coding, reasoning and vision; the model has a 1M-token context window and up to 128K output tokens.
How much does Claude Fable 5.2 cost?
On the API, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, 20% less than Fable 5.1 ($10 / $50). That is still twice the price of Claude Opus 5.5 ($4 / $20).
Is Claude Fable 5.2 included in Claude Pro?
On Pro, Fable models normally run on pay-as-you-go usage credits. As a launch offer, Anthropic includes them in the Pro plan and in Team standard seats from 9 to 16 November 2026 (until 11:59 pm PT), for up to 50% of weekly usage limits plus one Fable limit reset.
Fable 5.2 or Opus 5.5, which should a small company use?
Start with Claude Opus 5.5 for most work, which is also what Anthropic recommends. Use Fable 5.2 for demanding reasoning, long-horizon agentic work, or tasks where Opus 5.5 at higher effort still falls short.
What is the difference between Claude Fable 5.2 and Claude Mythos 5.2?
According to Anthropic, they are the same underlying model with the same specs and pricing. Fable 5.2 is generally available with cybersecurity and biology safeguards, while Mythos 5.2 is only for vetted organisations through the Cyber Verification Program and the Life Sciences Verification Program.