Skip to main content
  • Pricing
News

What Is Claude Fable 5? Fable 5 and 5.1 Explained

By Oakgen TeamPublished Updated 8 min read

Share
What Is Claude Fable 5? Fable 5 and 5.1 Explained

Claude Fable 5 is Anthropic's Mythos-class model for general use. It came out on June 9, 2026 with a 1M token context window and API pricing of $10 per million input tokens and $50 per million output tokens. On September 1, 2026 Anthropic released Claude Fable 5.1 at the same prices, and Fable 5 became a legacy model.

Everything below comes from Anthropic's and OpenAI's own pages and Oakgen's code, all read on 2026-10-25. We did not benchmark these models ourselves.

How do Fable 5, Fable 5.1, Mythos and Opus 5.5 compare?

The table uses Anthropic's model pages as of 2026-10-25. Prices are per million tokens on the Claude API.

ModelReleasedStatusInput / output priceCache readContext / max outputWho can use it
Claude Fable 5June 9, 2026Active (legacy), retirement not before June 9, 2027$10 / $50$11M / 128KEveryone
Claude Fable 5.1September 1, 2026Active (latest)$10 / $50$0.251M / 128KEveryone
Claude Mythos 5 and 5.1Alongside Fable 5 and 5.1LimitedSame as the matching FableSame as FableSame as FableVerified organizations only
Claude Opus 5.5Not checkedActive, retirement not before September 22, 2027$4 / $20$0.201M / 128KEveryone

Anthropic's models overview says to start with Opus 5.5 "for most workloads" and to use Fable 5.1 "for demanding reasoning and long-horizon agentic work", or when Opus 5.5 at higher effort still falls short in your own evaluations.

What is Claude Fable 5?

Fable 5 is the general-access version of Anthropic's Mythos-class model. The launch post opens with "Today we're launching Claude Fable 5: a Mythos-class model that we've made safe for general use." Anthropic places the Mythos class a tier above Opus.

The API model ID is claude-fable-5. It takes text and images and returns text. Thinking is adaptive and always on, with a default effort of high. Its reliable knowledge cutoff is January 2026, per the model page.

The "safe for general use" part is a set of safety classifiers. The launch post says they cover cybersecurity, biology and chemistry, and distillation. In Claude's apps, a flagged request is "automatically handled by Claude Opus 4.8 instead", and the user is told. Anthropic says the classifiers trigger "on average, in less than 5% of sessions" and that they can catch harmless requests.

When was Claude Fable 5 released, and why was it suspended?

Fable 5 launched on June 9, 2026, was pulled three days later and came back on July 1. The launch post's related updates give the dates. They do not give a reason for the suspension, so we do not repeat one here.

DateWhat happenedSource
June 9, 2026Fable 5 and Mythos 5 launched. Fable 5 went live on the API and in Claude appsLaunch post
June 12, 2026Anthropic suspended access to Fable 5 and Mythos 5Update linked from the launch post
June 23, 2026Fable 5 left Pro, Max, Team and seat-based Enterprise subscriptions and needed usage credits thereLaunch post
July 1, 2026"Claude Fable 5 and Mythos 5 are now available"Update linked from the launch post
September 1, 2026Fable 5.1 released. Fable 5 became legacyFable 5.1 model page

The lesson for builders is to keep a fallback model in your code. A model you depend on can lose access for days.

Where can you use Claude Fable 5 and 5.1?

Both are on the Claude API, Amazon Bedrock, Claude Models on Google Cloud, Microsoft Foundry and Claude Platform on AWS, per Anthropic's model pages. On Bedrock the IDs are anthropic.claude-fable-5 and anthropic.claude-fable-5-1. Everywhere else they match the Claude API IDs.

Inside Claude's own apps, plan access has changed since launch. Check the model picker in your account rather than a screenshot from June.

What changed in Claude Fable 5.1?

Fable 5.1 keeps Fable 5's input and output prices and cuts cache reads from $1 to $0.25 per million tokens. Anthropic's announcement says it beats Fable 5 on the benchmarks Anthropic reported, and that at low or medium effort it matches or beats Fable 5 at lower cost. The announcement estimates about 25% savings on typical workloads from the cheaper cache, and up to about 45% on heavily agentic work.

The safety changes matter for day-to-day use. Anthropic says cyber safeguards cause about 60% fewer interventions per Claude Code session, and Fable 5.1 can now help find software vulnerabilities, though not write exploits. Biology safeguards fire 85% less often on benign elementary biology and medical questions.

If you already call Fable 5, the 5.1 model page lists four breaking changes:

  • Forced tool use returns an error.
  • Only Fable 5.1 and Mythos 5.1 can read Fable 5.1's thinking blocks.
  • Editing earlier turns invalidates thinking blocks.
  • Some advisor tool pairings are not supported.

How is Claude Fable 5 different from Mythos 5?

They are the same model with different safeguards and different access. The launch post calls Mythos 5 "the same underlying model as Fable 5 but with the safeguards lifted in some areas". It started with Project Glasswing partners, in collaboration with the US government.

Today Anthropic's docs say Mythos 5 and 5.1 are only for organizations verified through programs such as the Cyber Verification Program. The 5.1 announcement adds a Life Sciences Verification Program and says access is limited to a set of US organizations. Specs and prices match the Fable models.

For almost every team, the answer is simple. Evaluate Fable 5.1. Plan around Mythos only if your Anthropic account team has confirmed access.

How much does the Claude Fable API cost?

Fable 5 and Fable 5.1 both cost $10 per million input tokens and $50 per million output tokens, as of 2026-10-25. The full price list for each:

Item, per million tokensFable 5Fable 5.1
Input$10$10
Output$50$50
5-minute cache write$12.50$12.50
1-hour cache write$20$20
Cache read$1$0.25
Batch API50% off input and output50% off input and output

At launch this was "less than half the price of Claude Mythos Preview", per the launch post. It is still the most expensive tier in Anthropic's current lineup. Opus 5.5 lists $4 and $20, and Sonnet 5.5 lists $2 and $10.

A sensible cost pattern is a router. Send drafting, formatting and extraction to a cheaper model. Send most production work to a mid-tier model. Keep Fable for hard code review, long documents and final checks where a wrong answer costs more than the call.

What happens when Claude Fable refuses a request?

On the API you get a normal response with stop_reason: "refusal", not an error. stop_details.category names the policy area. Anthropic's refusals page says you can usually still get an answer by sending the same request to another Claude model, and the API has a fallbacks parameter that does this for you.

Billing depends on when and why the refusal happened, per the same page as of 2026-10-25:

  • A refusal before any output is billed only in the bio, frontier_llm and reasoning_extraction categories. Other categories, or a null category, are not billed.
  • A refusal partway through a response bills the input and the output already streamed.
  • With fallback, the refusal is billed when it is mid-stream or in a billed category, plus the fallback request. A fallback credit covers the second prompt-cache write.

Plan the user experience for a refusal as a normal path. Tell the user, retry on an approved fallback when the task is safe, and avoid retry loops.

Is Claude Fable 5 good for coding?

Anthropic positions Fable 5.1 for "long-running agentic coding" and "multistep research". That points to work with history: multi-file migrations, architecture review before a risky change, debugging with logs and source together, and pull request review. For a one-function snippet, a cheaper model is usually enough.

Ask for diagnosis before edits. This prompt shape gives the model a job sequence:

You are reviewing a production bug. Do not edit yet.
1. Explain the likely failure path in plain English.
2. Name the smallest test that would reproduce it.
3. Propose the minimal fix.
4. List one thing that could break if the fix is too broad.

Then run the test yourself. A stronger model produces more convincing mistakes, so a failing test before the fix and a passing suite after it are still your proof.

How should business teams use Claude Fable 5?

Use it where the input is long and messy and the output is a decision. Good fits are turning long internal documents into a decision memo, comparing vendor proposals line by line, reading tables and charts alongside the narrative, and checking a policy for contradictions. Anthropic's 5.1 page also names document, spreadsheet and slide work.

Short emails, captions and simple formatting do not need it. A workable routine is:

  1. Paste or upload the source material.
  2. Have a cheaper model summarize and structure it.
  3. Ask Fable for the real decision, the trade-offs, the risks and what is missing.
  4. Ask a different model to critique that answer.
  5. Turn the final answer into a memo, slide outline or checklist.

How do you write good prompts for Claude Fable 5?

Give it a job description, not magic words. Anthropic publishes model-specific prompting guides for Fable 5 and 5.1, linked from each model page. This general structure separates the material from the instructions:

Context:
[What the model needs to know]

Task:
[The actual job]

Process:
1. Diagnose the situation.
2. State assumptions.
3. Give the answer.
4. Name risks or missing information.

Output:
[Format, length, tone and audience]

For a long document, add grounding rules:

Use the document as the source of truth.
If the document does not support an answer, say so.
Give section or heading references where you can.
End with the questions a reviewer should ask next.

For business analysis, ask for a decision, not "insights". Ask it to separate facts from interpretation, name the decision, give options, recommend one and state the strongest argument against it.

How does Claude Fable 5 compare with GPT-5.6?

We did not test them side by side. On published facts, GPT-5.6 is a three-model family from OpenAI and costs less per token. OpenAI's system card, published July 9, 2026, describes Sol as the flagship, Terra as "a capable lower-cost option" and Luna as the fastest and cheapest.

ItemClaude Fable 5.1GPT-5.6 SolGPT-5.6 TerraGPT-5.6 Luna
Input / output per million tokens$10 / $50$4 / $20$2 / $12$0.20 / $1.20
Context window1M1.05MNot checkedNot checked
Max output128K128KNot checkedNot checked

Prices come from Anthropic's model page and OpenAI's pricing page, both read on 2026-10-25. OpenAI's Sol page says its current price is promotional "at least through November 21, 2026", and it doubles input price for prompts over 272K tokens. OpenAI's models page now also lists GPT-6 models, so GPT-5.6 is no longer OpenAI's newest family. Our GPT-5.6 guide covers that family in detail.

What are the best alternatives to Claude Fable 5?

Within Anthropic, Opus 5.5 is the default Anthropic itself recommends, at $4 and $20 per million tokens. Sonnet 5.5 at $2 and $10 is the speed and cost balance, and Haiku 5.5 from $0.10 and $0.50 is for classification, extraction and routing.

Outside Anthropic, GPT-5.6 Sol and Terra cost less per token, and OpenAI's GPT-6 models are newer. Pick by job:

JobReasonable first choiceWhy
Hard multi-file code change or long-horizon agentFable 5.1, or Opus 5.5 at high effortAnthropic's own guidance for these models
Everyday drafting and analysisOpus 5.5 or Sonnet 5.5Lower price, same 1M context
High-volume classification or extractionHaiku 5.5 or GPT-5.6 LunaLowest listed prices
A second opinion on an important answerA model from a different vendorDifferent training, different blind spots

Can you use Claude Fable 5 on Oakgen?

No. Oakgen does not offer Fable as a model you can choose. Oakgen's Agent Chat offers Oakgen's own tiers: Oakgen Auto (the default), Oakgen Max and Oakgen Pro Max.

Agent Chat bills chat in credits per token. As of 2026-10-25, the per-token rate for Max is twice the rate for Auto, and the rate for Pro Max is five times the rate for Auto. Agent Chat needs an active Oakgen subscription, and it holds a credit reservation before each run, then charges for the tokens actually used. See pricing for plans. Agent Chat does not offer GPT-5.6, Fable or any other outside chat model as a named choice.

What Agent Chat adds over a plain chat window is creation. The same conversation can call Oakgen's image, video and audio models, so the memo you write with Pro Max can turn into a product image, a short video or a voiceover without switching apps.

Try Oakgen Agent Chat

Pick Oakgen Pro Max for hard reasoning, Max or Auto for everyday work, then create images, video and audio in the same chat.

Open Agent Chat

When is another tool a better choice?

  • If you need the API in your own product, call Anthropic, Bedrock, Google Cloud or Foundry directly. Oakgen Agent Chat is a workspace, not an API reseller.
  • If you only need chat and never generate media, a Claude subscription on claude.ai is the more direct route.
  • If your work is defensive security research that the Fable classifiers block, apply for Anthropic's Cyber Verification Program for Mythos access.
  • If you want GPT, Gemini or Grok answers, use those vendors' apps. Oakgen does not offer them in Agent Chat.

How we checked this

On 2026-10-25 we read Anthropic's Fable 5 launch post, the Fable 5.1 announcement, the models overview, the Fable 5 and Fable 5.1 model pages and the refusals and fallback page. We read OpenAI's API pricing page, the GPT-5.6 Sol model page and the GPT-5.6 system card page. Oakgen facts come from Oakgen's Agent Chat model list and its generated model price table, read the same day.

We did not run Fable 5, Fable 5.1, Mythos or GPT-5.6 side by side for this article, and nothing here is a benchmark result of ours. Benchmark claims are Anthropic's or OpenAI's own. Vendor prices and Oakgen credit prices can change, so check the vendor page and Oakgen's pricing page before you commit.

Frequently asked questions

Is Claude Fable 5 still available?

Yes. Anthropic's model page lists Claude Fable 5 as 'Active (legacy)' with retirement not sooner than June 9, 2027, checked 2026-10-25. Anthropic recommends migrating to Claude Fable 5.1, released September 1, 2026 at the same input and output prices.

What is the difference between Claude Fable 5 and Mythos 5?

Anthropic describes Mythos 5 as the same underlying model as Fable 5 with safeguards lifted in some areas. Fable 5 is open to all customers. Mythos 5 is only for organizations verified through Anthropic programs such as the Cyber Verification Program. Both share the same specs and pricing.

How much does Claude Fable 5 cost?

On the Claude API, Fable 5 and Fable 5.1 both cost $10 per million input tokens and $50 per million output tokens, per Anthropic's docs checked 2026-10-25. Cache reads cost $1 per million on Fable 5 and $0.25 on Fable 5.1. Batch requests are 50% off.

Can I use Claude Fable on Oakgen?

No. Oakgen does not offer Fable as a model you can choose. Agent Chat offers Oakgen's own tiers, Oakgen Auto, Oakgen Max and Oakgen Pro Max, as of 2026-10-25. It needs an active Oakgen subscription and bills chat in credits per token, with Pro Max at five times the per-token rate of Auto. To call Fable 5.1 directly, use the Claude API or a cloud platform that lists it.

Is Claude Fable 5 better than GPT-5.6?

We did not run a head-to-head test. On price, GPT-5.6 Sol lists $4 input and $20 output per million tokens on OpenAI's pricing page, against $10 and $50 for Fable. Pick by running your own task on both and checking the result.

Sources

  1. Anthropic, Claude Fable 5 and Claude Mythos 5 launch post
  2. Anthropic, Claude Fable 5.1 and Mythos 5.1 announcement
  3. Claude Platform docs, models overview
  4. Claude Platform docs, Claude Fable 5 model page
  5. Claude Platform docs, Claude Fable 5.1 model page
  6. Claude Platform docs, refusals and fallback
  7. OpenAI API docs, pricing
  8. OpenAI API docs, GPT-5.6 Sol model page
Share