Claude Fable 5 is Anthropic's top available model, and Claude Mythos 5 is the less restricted model. Both are available from June 2026, and Anthropic has patched them to new versions, Fable 5.1 and Mythos 5.1. This guide introduces you to the models, their costs, and how to access them via the API, as well as which one you should be interested in.
Key Importance:
Fable 5 and Mythos 5 were released on 9th June 2026. On 1st September 2026, Fable 5.1 and Mythos 5.1 were released.
The cost of Claude Fable 5.1 is $10 per million input tokens and $50 per million output tokens.
Its token context window is 1 million tokens and has up to 128K tokens of output.
According to Anthropic, Fable and Mythos are based on the same underlying model, but Fable has additional security measures.
Fable 5.1 is not available for free, but it is available for Pro, Max, Team and Enterprise users.
What is Claude Fable 5?
Claude Fable 5 is Anthropic's primary knowledge model for tasks like coding and hard knowledge work. It's between Opus, Sonnet and Haiku in terms of capability and cost. For long agent runs, large codebases, and complex analysis, an improved model is worth the additional expense, says Anthropic.
The current version is Claude Fable 5.1. If you read older articles about Fable 5, most of what they say still holds, with a few updates that are mentioned below.
Release Date and Timeline
The launch had an unusual schedule, so the dates are worth getting right:
June 9, 2026: Anthropic releases Claude Fable 5 and Claude Mythos 5.
June 12, 2026: Anthropic suspends access to both models to comply with U.S. Department of Commerce export controls.
June 30, 2026: The Department lifts those controls.
July 1, 2026: Anthropic restores access.
September 1, 2026: Claude Fable 5.1 and Claude Mythos 5.1 arrive.
Anthropic explains the suspension in its statement on Fable and Mythos access. Things can still change, so check the page for newer updates.
Claude Fable 5 vs Mythos 5: What Is the Difference?
Fable 5 is not a weaker, cheaper version of Mythos. According to Anthropic, the two share the same underlying model, but the difference is the safeguards. Fable adds extra protection in three areas: biology, cybersecurity, and research on training AI models. When a request touches those areas, Fable's safeguards step in.
Mythos is the version with those restrictions relaxed, so Anthropic handles access to it differently. Its first model, Claude Mythos Preview, is still limited to a small group of trusted organisations through Project Glasswing. You can read about that program on Anthropic's Glasswing page. For almost everyone else, Claude Fable 5.1 is the model to use.
Claude Fable 5 Price
Here is the current price list for Claude Fable 5.1 on the API, per million tokens:
Item | Price |
Input | $10 |
Output | $50 |
Cache read | $0.25 |
Cache write (5 minutes) | $12.50 |
Cache write (1 hour) | $20 |
Batch processing | 50% off input and output |
Fable 5.1 kept the same base prices as Fable 5. The one change is cache reads, which dropped 75 per cent from $1 to $0.25 per million tokens. Anthropic estimates this lowers the cost of typical workloads by about 25 per cent, and of heavy agent workloads by up to around 45 per cent. Confirm the numbers on Anthropic's Fable page before you decide the budget.
Two things can push your bill above the list price. Output tokens cost five times as much as input tokens, and the model's reasoning counts as output. Higher effort settings spend more reasoning tokens, so use the lowest effort that gets the job done.
How Fable 5.1 Compares on Price
Model | Input | Output |
Claude Fable 5.1 | $10 | $50 |
GPT-5.6 Sol | $5 | $30 |
Kimi K3 | $3 (uncached) | $15 |
Fable 5.1 is the most expensive option here. Anthropic's own Opus models list at roughly half the price of Fable, so for simpler tasks a cheaper Claude model is often the better deal. Our guides to Claude Opus 5, the GPT-5.6 review and Kimi K3 go deeper on the alternatives. If you are migrating from the previous generation, see our Claude Opus 4.8 review.
Claude Fable 5 API
The model ID is claude-fable-5-1. It has a 1 million token context window and a maximum output of 128,000 tokens, and the full window is billed at standard per-token rates, with no extra tier for long inputs. A bigger request still contains more billable tokens, so long contexts cost more simply because there is more to read.
Prompt caching matters a lot here. If your agent keeps re-reading the same large context, cache reads are $0.25 per million tokens; this makes Fable 5.1 far cheaper than the headline price suggests. Anthropic rates its latency as slower than the Opus models, which is worth knowing for real-time products.
Is Claude Fable 5 Free?
No. Anthropic lists Fable 5.1 as available to Pro, Max, Team and Enterprise users, and the API is paid per token. The free plan is not on that list. Limits and plan details change, so check your account before you count on it.
What Is Claude Fable 5 Good For?
Long coding and agent tasks: Large refactors, multi-file debugging, and agents that run for hours. Our guide to Claude Code for law firms and consultancies shows how teams use Claude in document-heavy work.
Big document sets: The 1M context window lets you load many contracts or reports into one prompt.
Hard analysis: Research synthesis, due diligence, and complex reasoning, where mistakes are expensive.
It is a poor fit for simple chatbot replies, high-volume classification, or anything where cost per call matters more than quality.
Which Claude Model Should You Choose?
Use Fable 5.1 when the task is hard, and the cost of a wrong answer is high. Use a cheaper Claude model when the task is routine. If you work with sensitive security topics, remember that Fable's safeguards may decline some requests that Mythos-level access would allow, so plan for that in your workflow.
In this guide to Claude Code for law firms and consultancies, we'll explore how teams are utilising Claude in a document-heavy style of work.
Many documents: You can load many contracts or reports into one prompt with the 1M context window.
Hard analysis: Research synthesis, due diligence, and complex reasoning if mistakes are costly.
It is not really suitable for just a simple chatbot response, a lot of classification, or any place where the expense for calls is more important than top quality.
Which Claude Model to use?
When the task is difficult and the consequences of a mistake are significant, use Fable 5.1. Utilise a less expensive Claude when the job to be done is routine. If you're working with a sensitive security topic, keep in mind that some requests are not allowed on Fable even if they would be allowed on Mythos level, so plan accordingly.
Conclusion
Claude Fable 5.1 is the strongest of the models you can use today and is priced accordingly by Anthropic. Choose it for its usefulness in tough work and for its cost control measures using caching. Before you proceed, review the current price and availability on Anthropic's pages.
Looking for guidance in selecting and implementing the ideal AI model for your team? Talk to TechNow, and we will help you compare, test and install the right solution for you.
Frequently Asked Questions
What is Claude Fable 5?
Anthropic’s generally available frontier model, released on June 9, 2026, and updated to Claude Fable 5.1 on September 1, 2026.
How much does Claude Fable 5.1 cost?
$10 per million input tokens and $50 per million output tokens, with cache reads at $0.25 per million tokens.
Is Claude Fable 5 free?
Fable 5.1 is listed in the Pro, Max, Team, and Enterprise versions of Anthropic, but not in the free variant.
What is the difference between Fable 5 and Mythos 5?
Anthropic claims that they are using the same base model, while Fable includes protection for biology, cybersecurity, and LLM research.
What was the reason for a ban on access to Fable 5?
On June 12, 2026, Anthropic temporarily suspended access to comply with U.S. export controls, and reopened access on July 1.
What is the context window size of Claude Fable 5.1?
1,000,000 tokens, with a maximum output of 128,000 tokens, at standard per-token pricing.