Claude Fable 5: pricing, capabilities and the most turbulent launch of the year
Anthropic's most intelligent model managed to pack a launch, benchmark record, White House-ordered shutdown and return to general availability into less than a month. I've gathered everything in one place: current pricing, hard numbers, the difference between Fable 5 and Mythos 5, and the complete timeline of this saga.
as of: 11 November 2026
Anthropic released the successor along with Mythos 5.1. Pricing hasn't changed ($10 and $50 per million tokens), context window also stayed at one million. Independent benchmarks haven't measured it yet, so for now your own tests are what counts.
The most expensive model on the market at twice the rate of Opus 5 ($5/25). In the Claude app it's included in paid plans; in Max without extra charges, beyond the plan limit works on usage credits. Response limit: 128k tokens. Anthropic requires 30-day data retention in the API for this model.
Timeline: from launch to return
Benchmarki
| Test | Fable 5 | Opus 5 | Opus 4.8 | co mierzy |
|---|---|---|---|---|
| SWE-bench Pro | 80.3% | 79.2% | 69.2% | coding on real repositories |
| Frontier-Bench v0.1 | 33.7% | 43.3% | 21.1% | tasks at the edge of model capabilities |
| GDPval-AA v2 | 1747 | 1861 | - | economically valuable work |
| Intelligence Index | 65/100 | - | - | highest score in the history of this test |
SWE-bench, Frontier-Bench and GDPval numbers come from Anthropic's launch materials (manufacturer claims, not independent measurement). The Intelligence Index score I describe in my review. A dash means no published result. Frontier-Bench results for Fable 5 and Opus 5 differ in methodology, so compare them carefully.
Fable 5 or Opus 5?
Fable 5 is the choice for tasks where only quality matters: the hardest code, multi-day agent work, document analysis on a million-token window. For everything else it makes more sense to pay half the rate for Opus 5, which is a hair's breadth away in most benchmarks. How much each costs in dollars for specific tasks, I calculate in cost calculator. You'll find the full review of the model in Fable 5 review.
Frequently Asked Questions
How much does Claude Fable 5 cost?
In the API: $10 per million input tokens and $50 per million output tokens, twice as much as Opus 5. In the Claude app it's part of paid plans; on the Max plan it's still included in the subscription without extra fees, beyond the plan limit you can purchase additional usage credits. I calculate costs for specific tasks in the AI cost calculator.
How does Fable 5 differ from Mythos 5?
It's the same model underneath. Fable 5 is the publicly available version with additional safeguards: classifiers catch risky queries in cybersecurity and biology areas, and such queries are handled by Opus 4.8. Mythos 5 has these safeguards removed and goes exclusively to verified organizations, including partners in the cyber defense program.
Is Fable 5 available worldwide?
Yes. After the June shutdown period, the model returned on July 1st to standard channels: Claude app, API (identifier claude-fable-5), and Claude Code. If you previously saw an access denied message, it should work normally now.
Is Fable 5 better than Opus 5?
In the toughest tasks yes: it leads in SWE-bench Pro (80.3 to 79.2 percent) and remains Anthropic's recommendation for multi-day agent work. Opus 5 wins on economics though: costs half as much and in many benchmarks is neck and neck. Fable 5 makes sense where response quality matters more than price and speed.
What is Fable 5's biggest weakness?
Speed and price. In deep reasoning mode, time to first response can exceed 100 seconds, and the token rate is the highest on the market. For simple, high-volume tasks it's overkill - a cheaper and faster model performs better there.
What is the context window of Fable 5?
Million tokens, with a 128 thousand token response limit. A single conversation can fit an entire novel along with project documentation. Large context window doesn't mean perfect memory across the full length though, worth testing with your own documents.
// follow
New tests and price changes, straight to your feed reader
Free RSS, no account and no email needed.
Follow Quantslant →Sources: Anthropic - launch announcement · model documentation · API pricing· own reports from the timeline above