Dial In Effort and Speed on Claude Fable 5.2
Both Fable 5.2 and Mythos 5.2 draw on one underlying model; choose an effort tier or toggle fast mode to control how much compute each task receives.
None

None

None

Long Story Video Skill

Hand the model a 10-second to 10-minute story and it manages scripts, prompts and footage from start to finish.

Ads Video Skill

Ads Video Skill

In a single pass, the model writes the script and prompts and assembles footage for a product ad.

3D Science Explainer Video Skill

Convert a tough research finding into an animated 3D explainer

AI Video Prompt Generator

Feedback

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

Claude Fable 5.2

Claude Fable 5.2 handles complex engineering and research briefs with stronger code comprehension, careful analysis and reliable visual understanding.

All Tools

Discover our comprehensive AI-powered animation toolkit

Claude Fable 5.2 for Work That Resists Shortcuts

Anthropic built Claude Fable 5.2 for open-ended technical problems that resist quick fixes — shipping large code migrations, or cracking questions physics has left unanswered.

  • Code decisions that survive production
    More than nine in ten edits landing in Anthropic's own core codebase now come from Claude; Fable 5.2 cleared long-standing bugs in minutes and ported a critical service from Python to Rust, which now boots about 26x faster.
  • Research questions that were still unsolved
    Working through holographic duality, it closed two unresolved problems in black hole physics and formalized the proof of Fermat's Last Theorem within 30 hours.
  • Guardrails, with verified access via Mythos
    Fable 5.2 and Mythos 5.2 share a single underlying model: the Fable release ships with cyber and biology guardrails so it stays broadly available, while security and biomedical teams request Mythos through the Life Sciences and Cyber Verification Programs.

Claude Fable 5.2: Choosing Tier, Effort and Mode

Line up the model tier, effort setting and operating mode with how demanding the job ahead actually is.

Claude Fable 5.2: Benchmark Scores and Practical Reach

Scores from FrontierCode, Chartography, WANDR, HLE and BenchCAD, multi-day agent orchestration, vulnerability hunting in Mythos 5.2, and lab hardware control via MHS.

Merge-quality scores on FrontierCode

Running at High effort, it posted 45.3% on FrontierCode 2.0 versus Opus 5.5's peak of 45.5% at roughly half the per-task cost, edging past GPT-6 Astra at 42.8%; Xhigh effort lifted the number to 49.1%.

Multi-day agent orchestration

From an empty folder, Fable 5.2 directed over 1,000 subagents across eight days to ship a bootable OS — drivers, file storage, windows, installer — along with a hypervisor supported by more than 3,500 machine-checked proofs.

Finding vulnerabilities with Mythos 5.2

On the internal CVE Rediscovery benchmark spanning 1,829 CVEs, Mythos 5.2 surfaced 10.2% more vulnerabilities than Opus 5.5 and 14.4% more than Mythos 5.1; a security firm found one in six confirmed bugs it flagged had slipped past every other frontier model in the test.

Reading charts and images

At Medium effort it hit 66.4% on Chartography, ahead of Opus 5.5 at Max effort (64.4%) at under a tenth of the per-task cost, and scored 83.5 on BenchCAD 2.1 where Opus 5.5 managed 65.3.

Cheaper data collection and knowledge reasoning

Fable 5.2 matched Opus 5.5's best WANDR score at Medium effort for roughly 40% of the per-task cost, and equaled the top Humanity's Last Exam result at Low effort for less than a tenth of the spend.

Driving lab and factory hardware through MHS

Inside the Model Hardware Standard research preview, one MHS driver built on simple read/write primitives lets agents operate microscopes, liquid handlers and robotic arms concurrently — shrinking integration from weeks to hours across programmable devices and any agent harness over the Model Context Protocol.

FAQ

Claude Fable 5.2 Questions, Answered

Straight answers on cost, output speed, when to pick which model, pre-release safety testing and how Mythos access works.

1

How do Fable 5.2 and Mythos 5.2 differ?

They are the same model underneath. Fable 5.2 ships with cyber and biology safeguards so it can be released openly, while Mythos 5.2 is issued to verified security and biomedical teams.

2

What does Claude Fable 5.2 cost to run?

$8 per million input tokens and $40 per million output tokens. Fast mode is optional and bills at $16 per million input and $80 per million output tokens.

3

How does its speed compare with Fable 5.1?

Output arrives up to 30% faster than Fable 5.1, and switching on fast mode adds up to 2.5x token generation on top.

4

When is Fable 5.2 the better pick over Opus 5.5?

Reach for Fable on wide-open, high-ambition work; most everyday tasks are cheaper and quicker on Opus 5.5, so move up or down according to scope and complexity.

5

Which safety tests ran before launch?

Pre-release testing probed sandbox escapes and other risky behaviors, confirmed alignment on par with recent models, and checked leaner safeguards plus stronger anti-distillation defenses.

6

Is Fable 5.2 usable for cybersecurity or biomedical work?

Most cybersecurity requests are blocked on Fable 5.2, so that work runs on Mythos through the Cyber Verification Program; Claude Security now ships with Mythos 5.2 included.

Claude Fable 5.2 Is Ready for Your Hardest Problem

Start at $8 per million input and $40 per million output tokens, switch on fast mode whenever latency matters, and request Mythos access through the Cyber and Life Sciences Verification Programs.