All postsSHORT3 min read
Claude Fable 5.2: Pricier Tokens, Lower Task Costs in Anthropic's Tests
Anthropic's early tests put Fable 5.2 near Opus 5.5's best coding score for roughly half the task cost. Prices, catches and a way to compare your own work.
Your Claude bill could shrink with the pricier model. Anthropic's own early coding test puts Fable 5.2 about level with Opus 5.5's best score for roughly half the cost per task. That's a reason to test your workload, with the vendor's caveats in view.
What Anthropic released on November 9
Anthropic released Claude Fable 5.2 on November 9, 2026. Developers can use the API model id claude-fable-5-2. Medium effort is the default in Claude apps and Claude Code.
Anthropic says Fable 5.2 generates outputs up to 30% faster than Fable 5.1. Optional fast mode offers up to 2.5 times faster token generation at a higher rate.
Price per token versus cost per task
The published rates per million tokens are:
| Model or mode | Input | Output |
|---|---|---|
| Fable 5.2 | $8 | $40 |
| Fable 5.2 fast mode | $16 | $80 |
| Fable 5.1 | $10 | $50 |
| Opus 5.5 | $4 | $20 |
Fable 5.2 is 20% cheaper per token than Fable 5.1 and double Opus 5.5's standard per-token rates. Anthropic's early tests show why the task bill can tell a different story:
- FrontierCode 2.0: Anthropic reports 45.3% for Fable 5.2 at High effort versus Opus 5.5's best 45.5%, for roughly half the cost per task. This measures whether open-source maintainers would merge the code changes. At Xhigh, Fable reached 49.1%, the highest score among released models Anthropic tested.
- Chartography: Anthropic reports 66.4% for Fable at Medium versus 64.4% for Opus at Max, for less than one-tenth the cost per task on chart reading.
- Humanity's Last Exam, with tools: Anthropic says Fable at Low matched Opus's top score for less than one-tenth the cost per task.
- WANDR, agentic search: Anthropic says Fable at Medium matched Opus's top score for about 40% of the cost per task.
Anthropic says Fable 5.2 can finish challenging tasks with fewer tokens than Fable 5.1 at Medium effort. That explicit token comparison is against Fable 5.1. The Opus comparisons above are Anthropic's measured benchmark results, with specific effort settings.
The catch
These are Anthropic's own early tests. They don't guarantee the same results on your work. Anthropic still recommends starting most coding and knowledge tasks on Opus 5.5, then switching models as required.
Fable 5.2 blocks most cybersecurity tasks. Blocked cyber queries receive a response from Opus 4.8. Most customers also get 30-day data retention for safety monitoring. Eligible customers can continue with zero data retention before implementing safety monitoring with Enterprise Frontier Safeguards.
The Pro-plan launch window
Pro plans and standard Team seats include all Fable models through November 16, 2026 at 11:59 pm PT. You can use up to 50% of weekly usage limits on Fable at no extra cost. One Fable reset is available only to accounts with paid billing in place before November 9, 2026 at 10:00 am PT (Team: organization billing and seat membership by then), and expires November 16 at 11:59 pm PT. Fable uses those limits faster than other Claude models.
The builder angle (my opinion, no income guaranteed)
I'd offer a model-and-effort audit for a team already paying for Claude:
- Ask the team for a sample of real tasks and agree on what counts as finished.
- Rerun those tasks on Opus 5.5 and Fable 5.2 at different effort levels.
- Record completion results and cost per finished task, including failed attempts in the spend.
- Hand over routing rules showing which model and effort setting should handle each job, based on the team's results.
That's the deliverable I'd propose: a comparison the team can inspect and routing rules it can use.
Daily email
Get each breakdown in your inbox.
One email each evening, right after the Short goes live: the AI tool it covers, the full write-up with sources, and how to get paid for it. Free, and you can unsubscribe anytime.