Claude Fable 5 is Anthropic’s most powerful
publicly available AI model, launched on June9

It is the first Mythos class model released
to the general public meaning it uses the same
underlying technology as Anthropic’s Mythos models
that shocked the cybersecurity world earlier this
year by finding thousands of zeroday vulnerabilities
in every major operating system and browser. For finance professionals,
the headline number is this: Fable 5 scored highest
of any model on Hebbia’s Finance Benchmark for
senior level reasoning, and trading firm IMC
confirmed it aced their evaluations across factual
lookup, conceptual reasoning, root cause analysis,
and expected value analysis . It costs
$10 per million input tokens and $50 per million
output tokens less than half the price of the
previous Mythos Preview model . Here
is here’s what matters.

What Claude Fable 5 Actually Is

Claude Fable 5 is a Mythos class model that
Anthropic has made safe for general public use . To understand why that matters, you
need to understand what “Mythos class” means.

In April 2026, Anthropic released Claude Mythos
Preview a model so capable at cybersecurity that
it could find and exploit zero day vulnerabilities
in every major operating system and every major
web browser when directed to do so ,
It found bugs that had been hiding in software
for decades, including a 27 year old vulnerability
in OpenBSD an operating system known primarily
for its security. Anthropic limited
Mythos Preview to about 150 organizations through
Project Glasswing, a collaboration with the US
government focused on cyber defense .

Fable 5 uses the same underlying model as Mythos

The difference is safeguards. Fable 5 comes
with classifiers that automatically redirect
queries about cybersecurity, biology, and chemistry
to Claude Opus 4.8 Anthropic’s previous
top tier model when those queries could be
misused . More than 95%
of Fable 5 sessions involve no fallback at all,
meaning most users experience full Mythoslevel
performance without any restriction.

The name itself tells the story. “Fable” comes
from the Latin fabula, meaning “that which is
told,” while “Mythos” is its Greek equivalent.
Same root. Same capability. Different risk
profile .

Why Finance Professionals Should Pay Attention

This is where Fable 5 gets interesting for anyone
working in finance, investing, or financial
technology.

On Hebbia’s Finance Benchmark a test designed
to measure senior level financial reasoning
Fable 5 scored higher than any other model,
including GPT5 variants and Google’s Gemini Ultra. The benchmark tests document based
reasoning, chart and table interpretation, and
complex problem solving. These are not toy
exercises. They are the kind of multi step
analytical tasks that senior analysts and
portfolio managers handle daily.

Trading firm IMC ran Fable 5 through their own
internal trading analysis evaluations and reported
that it “aced the evaluations nearly across the
board,” covering factual lookup, conceptual
reasoning, rootcause analysis, and expectedvalue
analysis For context, IMC is a
global market maker that processes billions of
dollars in trades. When they say a model aces
their evaluations, that means the model can
genuinely assist with real financial
decision making not just summarize Wikipedia
articles about finance.

The practical implications for finance
professionals:

Document analysis: Fable 5 can read and reason
over long financial documents — earnings reports,
SEC filings, prospectuses, loan agreements — and
extract insights with senior analyst level
accuracy .

Chart and table interpretation: Feed it a
complex financial chart or a multi tab
spreadsheet and it can interpret trends, spot
anomalies, and summarize findings.

Multi step research: Fable 5 can work
autonomously for longer than any previous Claude
model, making it suitable for deep research
projects that span days rather than minutes, but is this better than deepseek, Article of benchmark will be soon published head to head bookmark our site.

Expected value analysis: IMC’s evaluation
specifically tested expected value reasoning —
the core of investment decision making. Fable 5
passed .

If you work in finance and you have been waiting
for an AI model that can actually handle
professional grade analytical work, this is the
closest thing to it right now.

Software Engineering: Months of Work in a Day

For developers and fintech builders, Fable 5’s
software engineering capabilities are equally
impressive.

During early testing, Stripe reported that Fable 5
compressed months of engineering into days. In a
50 million line Ruby codebase one of the largest
private codebases in the world the model
performed a codebase wide migration in a single
day that would have taken a human team over two
months by hand.

On Cognition’s FrontierCode evaluation, which
tests whether models can pass difficult coding
tasks while meeting the standards of high quality
production codebases, Fable 5 scored highest among
all frontier models, even at medium effort .
This means it does not just write code that works —
it writes code that meets production standards,
with proper error handling, testing, and
documentation.

GitHub’s early testing confirmed the pattern.
Their team described Fable 5 as taking on “complex,
long horizon coding tasks with a level of autonomy
and reliability that exceeded previous benchmarks”(citation:2). The key phrase is “long horizon.”
Previous AI models could handle individual coding
tasks. Fable 5 can sustain a multi day coding
project, plan across stages, delegate to
sub agents, and check its own work (citation:2).

Cursor called it “the state of the art model on
CursorBench” and said it “opened up a class of
long horizon problems that were out of reach for
earlier models” (citation:2).

For fintech developers building AI powered
financial tools reconciliation systems, trading
algorithms, compliance automation — Fable 5
represents a meaningful jump in what is possible
with AI assisted development.

Vision: The Model That Sees and Rebuilds

Fable 5 is the new state of the art model for
vision tasks . This has direct
implications for anyone working with financial
documents, charts, or data visualizations.

What Fable 5 can do with vision:

Extract precise numbers from detailed scientific
and financial figures.

Rebuild a web application’s source code from
screenshots alone — meaning you can show it a
design and it generates working code .

Understand diagrams, charts, and tables nested
inside files and PDFs, improving work across
document heavy fields like finance, legal,
analytics, and architecture.

The most striking demonstration: previous Claude
models struggled to play the video game Pokémon
FireRed even with complex helper harnesses that
gave them additional tools and navigation aids.
Fable 5 beat the entire game using only raw game
screenshots no maps, no navigation aids, no
extra game state information, It
learned to play by looking, not by being told.

For finance, the practical applications are
immediate. Imagine scanning a 200 page bond
prospectus as images rather than extracting text.
Or showing Fable 5 a screenshot of a Bloomberg
terminal chart and asking it to explain the
pattern and calculate the implied volatility.
Or uploading a photograph of a handwritten
financial statement and having it digitize and
analyze the numbers. These are not theoretical
capabilities. They are available today.

Cybersecurity: The Double Edged Sword

Fable 5’s cybersecurity capabilities are both
its most impressive and most dangerous feature.
Understanding this tension is critical for anyone
in finance, where cybersecurity threats are
constant.

During testing with Claude Mythos Preview (the
unrestricted version), Anthropic found that the
model could identify and exploit zero day
vulnerabilities in every major operating system
and every major web browser (citation:4). The
exploits it constructs are sophisticated. In one
case, Mythos Preview wrote a web browser exploit
that chained together four vulnerabilities,
writing a complex JIT heap spray that escaped
both renderer and OS sandboxes (citation:4). It
autonomously obtained local privilege escalation
exploits on Linux by exploiting subtle race
conditions. And it wrote a remote code execution
exploit on FreeBSD’s NFS server that granted
full root access to unauthenticated users by
splitting a 20 gadget ROP chain over multiple
packets (citation:4).

The scale of improvement over previous models
is staggering. When testing Firefox JavaScript
engine exploits, Claude Opus 4.6 developed
working exploits 2 times out of several hundred
attempts. Mythos Preview developed working
exploits 181 times, and achieved register
control on 29 more (citation:4).

The UK’s AI Security Institute (AISI) conducted
independent evaluations and found that Mythos
Preview is the first model to solve “The Last
Ones” a 32 step corporate network attack
simulation spanning initial reconnaissance
through full network takeover, estimated to
require human experts 20 hours to complete. Mythos Preview solved it from
start to finish in 3 out of 10 attempts and
completed an average of 22 out of 32 steps
across all attempts. On expert level capture theflag
challenges, it succeeds 73% of the time no
model could complete these tasks at all before
April 2025 .

For financial institutions, this cuts both ways.
On defense, Fable 5 (through the Mythos 5 trusted
access program) can help banks, payment processors,
and fintech companies find and fix vulnerabilities
before attackers do. On offense, if these
capabilities ever reach malicious actors without
safeguards, the financial system faces a new
class of automated cyber threats.

This is why Anthropic deployed Fable 5 with
classifiers that redirect cybersecurity queries
to the less powerful Opus 4.8 ,
The same capabilities that make it an extraordinary
defender also make it a potential weapon.

Claude Mythos 5: The Unrestricted Version

While Fable 5 is available to everyone, Mythos 5
is restricted to a small group .

Claude Mythos 5 is the same underlying model as
Fable 5, but with cybersecurity safeguards lifted
in some areas . It is being deployed
through Project Glasswing in collaboration with
the US government as an upgrade to Claude Mythos
Preview . Anthropic plans to expand
access through a broader trusted access program
that allows cybersecurity organizations to apply
systematically .

Anthropic is also planning a separate trusted
access program for biology and chemistry, where
Fable 5’s biology safeguards would be lifted
while cyber safeguards remain in place. This
program will enroll researchers from life science
organizations spanning fundamental and
translational research .

The drug design results from Mythos 5 are
particularly striking. Anthropic’s internal
protein design experts found that Mythos 5
accelerated aspects of the drug design process
by approximately ten times . In one
study, the model — given protein design and
bioinformatics tools but no human assistance —
matched or beat skilled human operators on 14
protein targets, executing all tasks normally
completed by a scientist: choosing binding sites,
selecting and running protein design tools, and
recovering from failures along the way. Nine of
the 14 targets yielded strong candidates currently
under investigation .

Mythos 5 is also Anthropic’s “first model to
consistently produce novel, compelling scientific
hypotheses.” In blinded headto head comparisons
against Opus class models, Anthropic’s scientists
preferred Mythos 5’s molecular biology hypotheses
approximately 80% of the time, and several have
been advanced to experimental evaluation .
One hypothesis — a novel mechanism for an E. coli
protein — was independently corroborated by a
separate research lab working on the same
problem.

Claude Fable 5 vs Every Other Frontier Model

How does Fable 5 stack up against the competition?
Based on Anthropic’s benchmarks and early customer
feedback, it leads in several critical areas.

Capability	Claude Fable 5	Previous Best (Opus 4.8)	What Changed
Software engineering	State-of-the-art on Cognition FrontierCode, Stripe migration in 1 day	Capable but slower	Months of work compressed to days (citation:1)(citation:2)
Financial reasoning	Highest score on Hebbia Finance Benchmark (citation:1)	Strong but not leading	Senior-level analyst accuracy on document reasoning
Vision	State-of-the-art — rebuilt web app from screenshots, beat Pokémon with vision alone (citation:1)	Struggled with Pokémon even with harnesses	Minimal scaffolding needed
Cybersecurity	Full Mythos capabilities with safeguards (citation:1)	N/A — Mythos was restricted	Now publicly accessible with classifier guardrails
Autonomous work	Can work for days on multi-stage projects (citation:2)	Hours, not days	Planning, delegation, self-checking across stages
Memory	File-based memory improved performance 3x more than for Opus 4.8 (citation:1)	Good but limited	Better persistent context across long tasks
Scientific research	First model to produce novel, compelling hypotheses (citation:1)	Could assist but not generate	Drug design accelerated ~10x

Early customer feedback reinforces the benchmarks.
GitHub described Fable 5 as handling “complex,
long horizon coding tasks with a level of autonomy
and reliability that exceeded previous benchmarks”(citation:2). Cursor reported it opened up “a
class of long horizon problems that were out of
reach for earlier models” (citation:2). Cognition
called it “the strongest results of any Claude
model we’ve had the opportunity to test” (citation:2).

For finance specific use cases, the Hebbia Finance
Benchmark result and IMC’s trading analysis
evaluation are the most relevant data points.
No other publicly available model has demonstrated
this level of financial reasoning capability.

Pricing and How to Access It Right Now

Claude Fable 5 is available today. Here is
exactly what it costs and how to get it.

API Pricing

Input tokens: $10 per million

Output tokens: $50 per million

Prompt caching: 90% input token discount
(same as previous models)

US only inference: 1.1x pricing for input
and output tokens (citation:2)

This is less than half the price of Claude Mythos
Preview, which cost $25 per million input tokens.

Subscription Plans

Anthropic is rolling out access in stages:

From today through June 22: Fable 5 is included
on Pro, Max, Team, and seat based Enterprise
plans at no extra cost.

On June 23: Fable 5 will be removed from those
plans. Continued use will require purchasing
usage credits.

After that: Anthropic aims to restore Fable 5 as
a standard part of subscription plans when
sufficient capacity allows. No specific date
has been announced.

Where to Access It

Claude Platform: claude.ai

Claude API: claude fable 5 model name

Amazon Web Services

Google Cloud

Microsoft Foundry

Consumption based Enterprise plans (citation:2)

IMPORTANT: Anthropic expects demand for Fable 5
to be very high and difficult to predict.
If you want to test it for free, do it before
June 22. After that, it will require credits on
subscription plans or direct API usage.

Data Retention Change

Using Fable 5 requires 30 day data retention for
safety monitoring — this applies to all traffic
on Mythos class models, both first party and
third party surfaces ,
Anthropic states they will not use this data to
train new models, and all human access to the data
is logged with deletion guaranteed after 30 days
in almost all cases . This is a
change from previous Claude models and matters
for financial firms with strict data handling
policies.

Safeguards: How Anthropic Keeps Fable 5 Safe

Anthropic’s approach to Fable 5 safety is
unprecedented in the AI industry. Understanding
how it works helps you evaluate the model’s
reliability for professional use.

The Classifier System

Fable 5 uses separate AI systems — classifiers —
that detect potential misuse, including jailbreak
attempts, and prevent the main model from
responding on sensitive topics.
When the classifiers detect a request related to
cybersecurity, biology and chemistry, or
distillation attempts, the response is
automatically handled by Claude Opus 4.8 instead, Users are informed whenever this
happens and are not charged Fable 5 prices for
rerouted requests .

How Strong Are the Safeguards?

Anthropic ran an external bug bounty that produced
no universal jailbreaks in over 1,000 hours of
testing . External red teaming
organizations also failed to find universal
jailbreaks on long form agentic tasks .
One external partner found that Fable 5’s
safeguards against harmful cyber queries were the
most solid of any model tested, including Opus
4.8 and Opus 4.7. Fable 5 complied with zero
harmful single turn requests relating to planning
a cyberattack, exploit development, or defense
evasion — whether or not the requests used any
of 30 different public jailbreak techniques

The Three Protected Domains

Cybersecurity: Classifiers cover both
exploitation and offensive cyber tasks broadly.
Anthropic designed them to prevent uplift for
malicious actors in reconnaissance, discovery,
lateral movement, and exploit development.

Biology and chemistry: Classifiers block most
requests related to biology and chemistry,
falling back to Opus 4.8. Anthropic acknowledges
this is broader than ideal and plans to narrow
these safeguards as the trusted access program
expands.

Distillation: After identifying large scale
attempts to extract Claude’s capabilities to
train competing models in authoritarian countries,
Anthropic added classifiers to prevent
distillation of Fable 5’s abilities.

The Trade Off

Anthropic has been transparent that the safeguards
are tuned conservatively at launch. They will
sometimes catch harmless requests — triggering in
less than 5% of sessions on average .
For finance professionals, this means the
occasional query about financial risk modeling or
market analysis might get flagged and routed to
Opus 4.8 if the classifier interprets it as
potentially sensitive. Anthropic is working to
reduce these false positives as quickly as
possible.

What This Means for the AI Industry

Fable 5’s launch signals several shifts that
affect everyone in finance and technology.

Mythos Class Capabilities Are Now Public

The capabilities that Anthropic kept behind closed
doors since April are now available to any
developer or organization willing to pay $10 per
million input tokens.
For financial institutions, this means
AI assisted security auditing, document analysis,
and research capabilities that were previously
restricted to government partners are now
accessible through the API.

The Safeguard Precedent

Anthropic’s two tier model approach — Fable for
the public, Mythos for trusted partners — sets a
precedent for how frontier AI companies may
handle increasingly powerful models .
If this approach works, expect other AI labs to
adopt similar models. For financial regulators
evaluating AI risk, Anthropic’s approach offers
a concrete framework: same model, different
access levels based on risk profile and
organizational trust .

Price Compression Is Accelerating

Fable 5 costs less than half of Mythos Preview
despite being a stronger model ,
This follows a broader industry trend where
frontier capabilities become cheaper at an
accelerating rate. For fintech companies building
AI powered products, the cost of integrating
frontier level intelligence into financial
applications just dropped significantly.

Long Horizon Autonomous Work Is Here

The jump from “AI completes tasks” to “AI
completes multi day projects with planning,
delegation, and self checking” is the most
significant capability shift in this release(citation:2). For financial firms, this means
AI can now handle extended research projects,
complex document reviews, and multi stage
analytical workflows that previously required
junior analyst teams working for days or weeks.

Limitations: What Fable 5 Cannot Do

No model is perfect. Here is where Fable 5
falls short or remains untested.

It cannot bypass its own safeguards. If you
need cybersecurity capabilities for legitimate
defense work, you need Mythos 5 access through
Project Glasswing or the upcoming trusted access
program — Fable 5 will redirect those queries
to Opus 4.8.

Financial advice is not its purpose. While Fable
5 excels at financial reasoning and analysis, it
is not a licensed financial advisor. Its outputs
on Hebbia’s Finance Benchmark and IMC’s
evaluations demonstrate analytical capability,
not investment advice.

The 30 day data retention requirement may be a
dealbreaker for some financial firms. If your
compliance policies prohibit any third party
data retention, you may need to evaluate whether
Fable 5’s data handling aligns with your
regulatory requirements .

Capacity constraints are real. Anthropic has
been transparent that demand will likely exceed
supply. After June 22, Fable 5 will require
credits on subscription plans, and access may
be intermittent during high demand periods.

Biology and chemistry safeguards are overly
broad at launch. Anthropic acknowledges this
and plans to narrow them, but at launch,
approximately 5% of sessions may experience
false positive fallback to Opus 4.8.

It has not been independently tested for
financial compliance. While IMC and Hebbia
tested analytical capabilities, no independent
evaluation has specifically tested Fable 5’s
compliance with SEC, FINRA, or other financial
regulatory requirements for AI use in
financial services.

Frequently Asked Questions
(Implement FAQPage schema for all 6 questions)

FAQ 1:
Q: What is the difference between Claude Fable 5
and Claude Mythos 5?

A: They are the same underlying model. Fable 5
comes with safeguards that redirect sensitive
cybersecurity, biology, and chemistry queries to
Claude Opus 4.8. Mythos 5 has those safeguards
lifted in some areas and is only available to
trusted partners through Project Glasswing and
the upcoming trusted access program.

FAQ 2:
Q: How much does Claude Fable 5 cost?

A: $10 per million input tokens and $50 per
million output tokens. Prompt caching provides
a 90% discount on input tokens. It is less than
half the price of Claude Mythos preview.

FAQ 3:
Q: Is Claude Fable 5 free to use?

A: Until June 22, 2026, it is included at no
extra cost on Pro, Max, Team, and seat based
Enterprise plans. After June 23, it will require
purchasing usage credits on subscription plans

FAQ 4:
Q: Can Claude Fable 5 be used for financial
analysis?

A: Yes. Fable 5 scored highest of any model on
Hebbia’s Finance Benchmark for senior level
reasoning and aced IMC’s trading analysis
evaluations across factual lookup, conceptual
reasoning, root cause analysis, and expectedvalue
analysis . It is the strongest
publicly available model for financial analysis
tasks.

FAQ 5:
Q: Is my data safe with Claude Fable 5?

A: Using Fable 5 requires 30 day data retention
for safety monitoring across all surfaces. Anthropic
states this data is not used for model training
and is deleted after 30 days. All human access to
the data is logged. For sensitive financial data,
evaluate whether this data handling aligns with
your firm’s compliance requirements.

FAQ 6:
Q: How does Claude Fable 5 compare to GPT 5 for
finance?

A: On Hebbia’s Finance Benchmark — which tests
document based reasoning, chart interpretation,
and problem solving — Fable 5 scored higher than
any other model tested . IMC’s
independent trading analysis evaluation further
confirmed its strength in financial reasoning. For finance specific tasks, Fable 5
currently leads, though benchmark performance
should always be validated against your specific
use cases.

DISCLAIMER

This article is for informational
purposes only and does not constitute financial
advice. Model capabilities described here are
based on Anthropic’s published benchmarks and
early customer reports. Benchmark performance
does not guarantee results in specific use cases.
Evaluate AI tools based on your own requirements
and compliance obligations before integrating
them into professional workflows.

AUTHOR BIO BOX

AUTHOR: VixitAI Editorial Team
ROLE: AI & Finance Desk

BIO: The VixitAI editorial team covers the
intersection of artificial intelligence and finance
for American audiences. We test AI tools, analyze
regulatory developments, and break down complex
financial technology into actionable guidance.