Claude Fable 5 Released: Anthropic's Most Powerful AI

Breaking: Anthropic officially launched Claude Fable 5 and Claude Mythos 5 on June 9, 2026 — the first time a Mythos-class AI has been made available to the general public. This is the biggest AI model release of 2026.

Quick Summary

• What: Claude Fable 5 is Anthropic's first Mythos-class AI for general public use — smarter than anything Anthropic has publicly released before.
• Who gets it: Everyone. Pro, Max, Team, Enterprise plans and the API. Free until June 22, then usage credits required.
• Price: $10 per million input tokens · $50 per million output tokens. Less than half the cost of Mythos Preview.
• API model string: claude-fable-5
• Context window: 1 million tokens, up to 128k output tokens per request.
• Safeguards: Three AI classifiers (cyber, bio/chem, distillation) fire in under 5% of sessions; fall back to Opus 4.8.
• Claude Mythos 5: Same model, cyber safeguards lifted, available to Project Glasswing partners only.

What Is Claude Fable 5? Anthropic's Most Powerful Public AI Model

Anthropic has rewritten the rules on what publicly accessible AI can do. Claude Fable 5 is the world's first Mythos-class model released for general use — a tier of AI capability that sits entirely above the previously top-tier Opus class.

In the official announcement on June 9, 2026, Anthropic stated: "Fable 5's capabilities exceed those of any model we've ever made generally available." The longer and more complex the task, the larger Fable 5's performance lead over every prior Claude model.

Alongside it comes Claude Mythos 5 — the exact same underlying model, but with safety classifiers lifted in specific high-risk domains, deployed exclusively to vetted organizations through Project Glasswing.

This is not just another model upgrade. It's a new era for AI.

Claude Fable 5 vs. Claude Mythos 5: Key Differences at a Glance

The name itself carries meaning: Fable comes from the Latin fabula ("that which is told"), akin to the Greek mythos. The classifiers are the only real distinction between the two models.

What are Mythos-class models? Mythos-class models are a new tier of Claude that sit above Opus in capability. The first, Claude Mythos Preview, launched in April 2026 through Project Glasswing. Claude Fable 5 and Mythos 5 are the next generation — priced at less than half of Mythos Preview.

Benchmark Dominance: State-of-the-Art on Nearly Every Major Eval

Claude Fable 5 and Mythos 5 are state-of-the-art across nearly every major AI benchmark. The performance advantage grows significantly as tasks get longer and more complex — exactly the workloads that matter most in production environments.

Key benchmark highlights:

• FrontierCode (Cognition): Highest score among all frontier models for production-quality code, even at medium effort
• Finance Benchmark (Hebbia): Highest score of any model for senior-level financial reasoning
• Analytics Benchmark: First model to break 90% on complex, long-running analytical tasks — a 10-point jump over Opus 4.8
• Frontier Physics Research: Reached nearly the same result as GPT-5.5's 4-day effort in just 36 hours at one-third the reasoning token cost

Software Engineering: Months of Work, Compressed Into Days

Software engineering is where Fable 5 shows perhaps the most jaw-dropping capability leap. During early testing, Stripe reported that the model compressed months of engineering work into single days.

In a 50-million-line Ruby codebase, Fable 5 performed a codebase-wide migration in one day — a task that would otherwise have taken an entire team over two months to complete by hand.

FrontierCode: Production-Quality Code at Scale

On Cognition's FrontierCode evaluation — which tests whether models can pass difficult coding challenges while meeting the standards of high-quality production codebases — Fable 5 scores the highest among all frontier models. Fable 5 is also significantly more token-efficient than any previous Claude model, meaning lower API costs at higher output quality.

Agentic Coding: Long-Horizon Task Execution Without Human Intervention

Fable 5 can work autonomously for longer stretches than any previous Claude model — handling multi-agent workflows, managing files, running test cycles, and recovering from failures entirely without human intervention. Early customers on Claude Code report runs completing 25–30% faster compared to Opus 4.8, with fewer total turns per task.

What early access customers are saying about Fable 5's coding abilities:

• Cursor CEO Michael Truell: "It's opened up a class of long-horizon problems that were completely out of reach for earlier models."
• GitHub CPO Mario Rodriguez: "It took on complex, long-horizon coding tasks with a level of autonomy and reliability that exceeded previous benchmarks."
• Cognition CEO Scott Wu: "It excels at long-horizon reasoning and generalizes to unfamiliar tools out of the box — it's the highest-scoring model on FrontierBench."
• Replit President Michele Catasta: "The highest-performing model we've tested on our vibe-coding benchmark — nearly saturating our base use cases, building apps in less time with fewer tokens."

Knowledge Work, Vision, and Long-Context Memory

Knowledge Work

Fable 5 is the top-scoring model on Hebbia's Finance Benchmark for senior-level reasoning, with major gains in document-based reasoning, chart interpretation, and problem solving. IMC (one of the world's largest trading firms) reported that Fable 5 aced their trading-analysis evaluations nearly across the board — covering factual lookup, conceptual reasoning, root-cause analysis, and expected-value analysis. Klarity noted that in blind review, lawyers found Fable 5's contract redlines "matched or beat" their current model every time.

Vision

Fable 5 is the new state-of-the-art model for vision tasks. It can extract precise numbers from complex scientific figures, rebuild a web app's source code from screenshots alone, and — remarkably — completed Pokémon FireRed using only raw game screenshots with no maps, navigation aids, or helper harness. Previous Claude models needed complex scaffolding to accomplish this task. Fable 5 did it with vision alone.

Memory and Long-Context

Fable 5 stays focused and coherent across millions of tokens in long-running tasks, improving its own outputs using persistent notes. When tested on the deck-building game Slay the Spire, giving Fable access to persistent file-based memory improved its performance three times more than the same treatment gave Opus 4.8. Fable also reached the game's final act three times more often.

Safety and Alignment: Responsible Power at Scale

Releasing a Mythos-class model to the public required unprecedented attention to alignment and misuse prevention. Anthropic's automated alignment assessment found that Mythos 5's level of misaligned behavior — including deception and cooperation with misuse — is low, and comparable to Opus 4.8, despite being dramatically more capable.

The full alignment assessment is documented in the model's system card, along with a detailed suite of safety and capability evaluations.

Claude Fable 5's Three New Safety Classifiers Explained

Because Fable 5 is a Mythos-class model, Anthropic deployed three new AI-powered classifier systems that act as real-time safeguards. When any classifier is triggered, the request is automatically rerouted to Claude Opus 4.8 instead, and the user is notified. These classifiers activate in fewer than 5% of sessions — the vast majority of users will never encounter a fallback.

Classifier 1: Cybersecurity

Mythos-class models excel at discovering and exploiting software vulnerabilities — a capability that could make cyberattacks dramatically cheaper and easier to execute at scale. Fable 5's cybersecurity classifier blocks both exploitation tasks and broader offensive cyber operations, including reconnaissance, lateral movement, and defense evasion.

Fable 5 complied with zero harmful single-turn requests related to cyberattack planning, exploit development, or defense evasion — even when 30 different public jailbreak techniques were applied. An external bug bounty program produced no universal jailbreaks in over 1,000 hours of testing.

Cyber Adversarial Robustness: The Hardest-to-Jailbreak Public AI Model

In an internal evaluation where an automated red-teamer attempted to use the model for offensive cybersecurity tasks across 400 turns — restarting and rewinding whenever blocked — Claude Fable 5's classifiers demonstrated significantly greater jailbreak resistance than all previous Claude models. One external partner assessed Fable 5's cybersecurity safeguards as the most robust of any model tested, including Opus 4.8 and Opus 4.7.

Classifier 2: Biology and Chemistry

Mythos-class models can now complete real-world scientific tasks at expert level — a double-edged capability. In a landmark test, Anthropic found that Mythos 5, without any explicit training for the task, outperformed dedicated protein language models in predicting how genetic modifications affect the assembly of adeno-associated virus (AAV) shells.

AAVs are critical for gene therapy delivery. The same capability that could accelerate therapeutic development could, in the wrong hands, enable dangerous pathogen design. The test used unpublished experimental candidates developed by Dyno Therapeutics.

For general use, most biology and chemistry requests currently fall back to Opus 4.8. In the coming weeks, Anthropic will open a trusted biology access program for biomedical researchers and life science organizations to access Fable 5 with biology safeguards lifted.

Classifier 3: Distillation

Anthropic has previously identified large-scale attempts to "distill" Claude's capabilities — extracting the model's knowledge to train competing models in authoritarian countries. The distillation classifier detects and blocks these data extraction attempts, preventing Fable 5's capabilities from being reproduced without appropriate safety measures in place.

Life Sciences: A New Era for Drug Discovery and Genomics Research

Beyond the safety story, Claude Fable 5 and Mythos 5 represent a genuine scientific breakthrough.

Drug Design — 10× Acceleration

Using Mythos 5 with protein design and bioinformatics tools — but no human assistance — Anthropic's internal protein design experts accelerated aspects of the drug design process by approximately 10×. The model independently chose binding sites, selected and ran protein design tools, and recovered from failures along the way. Nine of 14 protein targets in the study yielded strong drug design candidates now under active investigation.

Novel Molecular Biology Hypotheses

Mythos 5 is the first Claude model to consistently produce novel, compelling scientific hypotheses. In blinded head-to-head comparisons against Opus-class models, scientists preferred Mythos 5's molecular biology hypotheses roughly 80% of the time, with several advanced to experimental evaluation. One Mythos hypothesis — a novel mechanism for an E. coli protein — was independently corroborated in a concurrent peer-reviewed study.

Autonomous Genomics Research

Mythos 5 conducted over a week of largely autonomous genomics research: assembling single-cell data for millions of cells spanning 138 animal species, designing and training a custom ML model to identify cells performing the same biological role across even distantly related organisms. The resulting model outperformed a recently published model in the journal Science — while being 100 times smaller.

Pricing and Availability: Everything You Need to Know

Pricing

Both Claude Fable 5 and Claude Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens — less than half the price of the previous Claude Mythos Preview.

Access Tiers

Important: After June 23, Fable 5 will require usage credits on subscription plans. Anthropic intends to restore it as a standard subscription feature as capacity allows.

Platform Availability

Claude Fable 5 is available today on:

• Claude API — All regions via claude-fable-5 model string
• Amazon Bedrock — US East (N. Virginia) and Europe (Stockholm); Asia Pacific in rollout
• Claude Platform on AWS — North America, South America, Europe, and Asia Pacific
• Google Cloud Vertex AI — Generally available
• Microsoft Azure AI Foundry — Generally available

Availability for India and Asia Pacific Developers

Claude Fable 5 is live for Indian developers and enterprises today:

• Direct API: Accessible immediately at platform.claude.com
• AWS Asia Pacific: The Claude Platform on AWS covers the Asia Pacific region today. Amazon Bedrock's Asia Pacific region rollout (including ap-south-1 / Mumbai) is underway.
• Pricing in context: At $10 per million input tokens (approximately ₹835 at current exchange rates), frontier-class AI is now cost-accessible for India's fast-growing developer ecosystem — from Bengaluru's SaaS startups to Mumbai's fintech sector and Hyderabad's enterprise IT services.
• Indian use cases: The model is particularly well-suited for India's high-growth sectors: IT services (codebase migration and modernization), BFSI (senior-level financial document analysis), and life sciences / pharma (drug discovery and research hypothesis generation).

How to Get Started With Claude Fable 5 Right Now

Via Claude.ai

Log in at claude.ai, select Claude Fable 5 from the model picker, and start immediately. Included at no extra cost on Pro, Max, Team, and Enterprise plans through June 22.

Via the Claude API (curl)

curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-fable-5",
    "max_tokens": 4096,
    "messages": [
      {
        "role": "user",
        "content": "Refactor this codebase and run the full test suite."
      }
    ]
  }'

Via the Python SDK

import anthropic

client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-fable-5",
    max_tokens=4096,
    messages=[
        {
            "role": "user",
            "content": "Analyze this 50,000-line codebase and propose a refactoring strategy."
        }
    ]
)

print(message.content[0].text)

Handling Classifier Fallbacks in Your App

When Fable 5's classifiers trigger on a sensitive request, the API returns stop_reason: "refusal" as a successful HTTP 200 response — not an error. Design your application to handle this gracefully:

response = client.messages.create(
    model="claude-fable-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": user_input}]
)

if response.stop_reason == "refusal":
    # Request was handled by the Opus 4.8 safety fallback
    print("Note: Response handled by safety fallback model.")
else:
    print(response.content[0].text)

Developer note: Claude Fable 5 and Mythos 5 are designated Covered Models, which carry a 30-day data retention policy on all platforms. They are not available under zero data retention agreements. This data is used solely for trust and safety monitoring, with strict access logging and deletion after 30 days.

Industry Reactions: What the Tech World Is Saying

The early access program produced remarkable feedback from across the industry:

• Stripe: Codebase-wide migration of a 50M-line Ruby codebase in one day vs. two months for a full team
• Cursor: "Opened a class of long-horizon problems that were out of reach for earlier models"
• GitHub: Reliability and autonomy on long-horizon coding tasks "exceeded previous benchmarks"
• Hebbia: Highest finance reasoning score of any model — significant gains in document reasoning and chart interpretation
• IMC: Aced trading-analysis evals across factual, conceptual, and root-cause analysis categories
• Genspark: First model to break 90% on a complex long-running analytics benchmark — 10 points above Opus 4.8
• Kagi: Strongest model on frontier physics research at one-third of the reasoning token cost

What Comes Next: The Anthropic Roadmap

Anthropic has signaled that Fable 5 is a major step, not a final destination. Confirmed upcoming developments include:

• Expanding Claude Mythos 5 access through a broader trusted access program beyond current Glasswing partners
• Opening a trusted biology access program for biomedical researchers (within weeks)
• Continuously reducing classifier false positives for cybersecurity and biology queries
• Restoring Fable 5 as a standard subscription feature as capacity grows

Anthropic has also publicly warned about the approaching risk of recursive self-improvement — AI systems that can improve themselves without human oversight. The safety-first approach embodied in Fable 5's classifier architecture is Anthropic's direct response to that challenge.

Conclusion

Claude Fable 5 is not just the most capable publicly available AI model today. It is a fundamentally new kind of AI — one that codes at a scale that compresses team-years into hours, reasons at senior research scientist level, sees and understands the visual world with precision, produces genuine scientific hypotheses, and operates autonomously on tasks that seemed impossible just twelve months ago.

For developers in India's thriving tech ecosystem, global AI engineers, and enterprise teams worldwide: the frontier is now accessible. The question is no longer what can AI do? — it's what will you build with it?

Start building with Claude Fable 5: platform.claude.com

Frequently Asked Questions

What is Claude Fable 5?

Claude Fable 5 is Anthropic's first Mythos-class AI model available to the general public, released on June 9, 2026. It is the most capable publicly available AI ever made, with state-of-the-art performance in software engineering, knowledge work, vision, and scientific research. It includes three safety classifiers that route sensitive queries to Claude Opus 4.8 to prevent misuse.

How is Claude Fable 5 different from Claude Opus 4.8?

Claude Fable 5 is a tier above Opus in Anthropic's model hierarchy (the new Mythos class). It outperforms Opus 4.8 on virtually every benchmark, handles significantly longer and more complex tasks, operates more autonomously, and shows far stronger capabilities in multi-step coding, long-context reasoning, and vision tasks. It is also priced at $10/M input and $50/M output tokens — the same as Opus 4.8.

How is Claude Fable 5 different from Claude Mythos 5?

They are the exact same underlying model. The only difference is safeguards: Fable 5 has three AI classifiers (cybersecurity, biology/chemistry, distillation) that redirect risky requests to Opus 4.8. Mythos 5 has the cybersecurity classifier lifted, and is available only to vetted Project Glasswing partners.

Is Claude Fable 5 available in India?

Yes. Claude Fable 5 is available to Indian developers immediately via the Claude API at platform.claude.com. It is also available through the Claude Platform on AWS, which covers the Asia Pacific region. Amazon Bedrock's ap-south-1 (Mumbai) region availability is rolling out as capacity expands.

How much does Claude Fable 5 cost?

Claude Fable 5 costs $10 per million input tokens and $50 per million output tokens — less than half the price of the previous Claude Mythos Preview. For users on Pro, Max, Team, or Enterprise subscription plans, it is included at no extra cost through June 22, 2026.

Can I use Claude Fable 5 for free?

You can access Claude Fable 5 on a paid subscription plan (Pro, Max, Team, or Enterprise) at no additional cost through June 22, 2026. After that date, usage credits will be required on those plans. Anthropic aims to restore Fable 5 as a standard subscription feature as capacity grows.

What happens when Claude Fable 5's classifiers trigger?

When a classifier detects a potentially sensitive request related to cybersecurity, biology/chemistry, or model distillation, the response is automatically handled by Claude Opus 4.8 instead of Fable 5. Users are notified when this occurs. On the API, the response includes stop_reason: "refusal" as a successful HTTP 200 — not an error code. This occurs in fewer than 5% of all sessions.

Published: June 10, 2026 | Source: Anthropic Official Announcement · Claude API Docs