Claude Fable 5.2 vs Fable 5.1: Benchmarks, Pricing and Which to Use (2026)

Claude Fable 5.2 beats Fable 5.1 on every benchmark Anthropic reports, runs up to 30% faster and costs 20% less. Here's the full comparison, with pricing, a migration checklist and when to use Opus 5.5 instead.

Rana AqibNovember 22, 202613 min read
Claude Fable 5.2 vs Fable 5.1: Benchmarks, Pricing and Which to Use (2026) — HyzenPro article cover

Claude Fable 5.2 is the newest Fable model, and it beats Fable 5.1 on every benchmark Anthropic reports. It also costs 20% less ($8 / $40 per million tokens instead of $10 / $50) and writes output up to 30% faster.

If you're on Fable 5.1, upgrade. Just run one code check first, because a few API behaviors changed. If you're choosing between Fable and other Claude models, there's a twist: Anthropic itself says most people should start with Claude Opus 5.5 and only move up to Fable for the hardest work.

This guide covers both models, plus the Fable 5 launch that started the series. You'll get benchmarks, real prices, a migration checklist, and a clear way to decide which model to use.

Quick answer: Fable 5.2 vs Fable 5.1

  • Newer and stronger: Fable 5.2 (released November 9, 2026). Fable 5.1 came out September 1, 2026.

  • Price: Fable 5.2 is $8 / $40 per million tokens. Fable 5.1 is $10 / $50.

  • Speed: Fable 5.2 generates output up to 30% faster, and fast mode adds up to 2.5x more.

  • Safeguards: Fable 5.2 blocks about 55% fewer requests than Fable 5.1, with the same level of protection.

  • Best default for most people: Claude Opus 5.5, then step up to Fable 5.2 when you hit its limits. See where each sits on our AI leaderboard.

What is Claude Fable?

Claude Fable is Anthropic's top tier for the hardest technical work: long-running coding, research, finance and science. Anthropic describes Fable 5.2 as its most capable model for technical work.

Fable has a sibling called Mythos. They're the same underlying model with different safeguards. Fable is open to everyone. Mythos is only for vetted cybersecurity and life-science organizations. Want the full story on that split? Read our Claude Fable 5 vs Mythos 5 comparison.

The Fable timeline

Model

Released

API price (in / out per 1M)

Status

Claude Fable 5

June 9, 2026

$10 / $50

Legacy

Claude Fable 5.1

September 1, 2026

$10 / $50

Legacy

Claude Fable 5.2

November 9, 2026

$8 / $40

Current

The launch had a rough start. Three days after release, Anthropic suspended Fable 5 and Mythos 5 to comply with US export controls. Access returned July 1. You can read Anthropic's Fable and Mythos access statement and its note on restoring Fable 5.

Fable 5.2 vs Fable 5.1 at a glance

Spec

Claude Fable 5.2

Claude Fable 5.1

Released

Nov 9, 2026

Sep 1, 2026

API model ID

claude-fable-5-2

claude-fable-5-1

Input / output price

$8 / $40 per 1M

$10 / $50 per 1M

Cache reads

$0.20 per 1M

$0.25 per 1M

Context window

1M tokens

1M tokens

Output speed

Up to 30% faster than 5.1

Baseline

Fast mode

Up to 2.5x faster at $16 / $80

Not listed

Safeguard blocks

About 55% fewer than 5.1

Baseline

Knowledge cutoff

June 2026

Not listed

BenchLM overall score

93.3 (#1)

78.7 (#12)

Both need 30-day data retention by default for safety monitoring. Fable 5.2 also keeps a 128K max output and always-on adaptive thinking.

Benchmarks: Fable 5.2 vs Fable 5.1 vs Opus 5.5

These are Anthropic-reported scores from the Fable 5.2 announcement, at each model's highest available effort and with safeguards on. They aren't independent tests, so treat them as a guide.

Benchmark

Fable 5.2

Fable 5.1

Opus 5.5

FrontierCode 2.0 (agentic coding)

49.1%

32.4%

45.5%

FrontierSWE v2 (agentic coding)

69.9%

56.3%

62.3%

AutomationBench (business workflows)

51.4%

31.4%

42.5%

Terminal-Bench-Science 0.1

74.6%

52.6%

58.7%

WANDR (agentic search)

82.1%

68.7%

72.3%

OSWorld 2.1 (computer use, partial)

89.3%

83.5%

87.2%

RoboDojo Sim (robotics)

64.9%

27.6%

39.4%

Remote Labor Index

41.3%

17.9%

21.3%

Humanity's Last Exam (with tools)

74.2%

65.6%

67.7%

Chartography (chart reading)

73.4%

44.8%

64.4%

The jumps that stand out are robotics (64.9% vs 27.6%), the Remote Labor Index (41.3% vs 17.9%) and chart reading (73.4% vs 44.8%).

Where Fable 5.2 doesn't win

It isn't first everywhere. In Anthropic's own table, Gemini 4 Carbon scores higher on AutomationBench (53.9% vs 51.4%). Opus 5.5 has a slightly higher Elo on the AA-Briefcase knowledge-work test (1800 vs 1785), though Fable leads that test's rubric score (66.1% vs 61.2%).

For an outside view, BenchLM's independent index puts Fable 5.2 first at 93.3, ahead of Gemini 4 Neon (91.3) and GPT-6.2 Astra (90.8). Fable 5.1 sits at 78.7 and is now marked superseded. See our full top 20 frontier AI models ranking for the whole field.

What actually improved from Fable 5.1 to 5.2

Better results for the money

Price isn't the only saving. Anthropic reports that Fable 5.2 needs fewer tokens for the same work:

  • On FrontierCode 2.0 at High effort, it scores level with Opus 5.5's best result for roughly half the cost per task.

  • On the WANDR agentic search test at Medium effort, it matches Opus 5.5's top score for about 40% of the cost.

  • On Humanity's Last Exam at Low effort, it matches Opus 5.5's top score for under one-tenth of the cost.

Longer, bigger agent runs

Anthropic calls Fable 5.2 its strongest lead agent. Two of its examples show the scale:

  • One engineer had it design a prototype operating system from an empty folder. It coordinated over 1,000 subagents across eight days and produced a system that boots on a laptop.

  • A team of 1,000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in 30 hours. Opus 5.5, with the same setup, completed only 26% in 18 hours. Anthropic explains the setup in its multi-agent write-up.

Stronger coding judgment

Anthropic says over 90% of the changes merged into its core codebase are now written by Claude with human guidance. A team of agents running an early Fable 5.2 migrated a production service from Python to Rust, and the service now starts about 26 times faster.

Customers say similar things. Vercel's engineer said they never had to say "No, I meant..." once. Cognition noted that it admits when it doesn't know something instead of guessing.

Science, hardware and robotics

Fable 5.2 solved two open problems in theoretical physics about black holes. In a hardware test, it designed three working electronic devices from a one-paragraph prompt, as shown in Anthropic's electronics design write-up. On RoboDojo, it scored 64.9%, more than double GPT-6 Astra (29.8%).

Fewer unnecessary blocks

Fable 5.2's safety classifiers fire about 55% less often than Fable 5.1's: roughly 60% fewer in cybersecurity and 40% fewer in biology. The protection is as strong as before. It simply flags less harmless work.

What Fable 5.1 added over Fable 5

Fable 5.1 is worth a look if you still run it or want the series history. Anthropic's Fable 5.1 announcement lists these upgrades over Fable 5:

  • Lower cost. Cache reads dropped 75% to $0.25 per million tokens. Typical workloads cost about 25% less, and heavy agentic work up to about 45% less.

  • Better results. Terminal-Bench-Science 0.1 rose from 24.7% to 52.6%. Terminal-Bench 4.0 went from 42.0% to 55.8%. AutomationBench nearly doubled, from 17.1% to 31.4%.

  • Fewer false blocks. Cyber safeguards produced about 60% fewer false positives, and the model could now help find software vulnerabilities (though not build exploits).

  • Enterprise privacy. It introduced Enterprise Frontier Safeguards, which let enterprises get zero-data-retention-level privacy while still allowing safety monitoring.

One example shows the leap. Investment firm Millennium said Fable 5.1 found the cause of a roughly one-in-a-million crash that its engineers hadn't explained in four to five years.

Pricing: Fable 5.2 vs Fable 5.1 vs Fable 5

Price per 1M tokens

Fable 5.2

Fable 5.1

Fable 5

Input

$8

$10

$10

Output

$40

$50

$50

Cache reads

$0.20

$0.25

$1

Batch API (input / output)

$4 / $20

Not listed

50% off

Fast mode (input / output)

$16 / $80

Not listed

Not listed

Fable 5.2 also offers US-only inference at 1.1x pricing. Check Anthropic's pricing page for current numbers.

A worked example. Say an agent run reads 20M cached tokens, 2M fresh input tokens and writes 500K output tokens.

  • On Fable 5.2: $4.00 + $16.00 + $20.00 = $40

  • On Fable 5.1: $5.00 + $20.00 + $25.00 = $50

Same tokens, 20% less. In practice Fable 5.2 may use fewer tokens too.

Fable 5.2 vs Opus 5.5 vs Sonnet 5.5: which should you use?

Model

Price (in / out per 1M)

Speed

Best for

Claude Fable 5.2

$8 / $40

Moderate

The hardest, longest, most ambiguous work

Claude Opus 5.5

$4 / $20

Moderate

Daily coding and knowledge work

Claude Sonnet 5.5

$2 / $10

Fast

Everyday tasks and high-volume steps

Claude Haiku 5.5

From $0.10 / $0.50

Fastest

Classification and simple jobs

Anthropic's advice is clear. Start with Opus 5.5. Move up to Fable 5.2 only when your tests on Opus at higher effort still fall short. Fable can also act as an advisor that smaller models consult, or as an orchestrator that hands simple tasks to cheaper models. That mix keeps quality high without paying Fable prices on every call.

Curious about the mid-tier? Read our Claude Sonnet 5 review. Or see how Claude's mid-tier stacks up for coding on our list of AI coding tools.

Pick Fable 5.2 if you:

  • run agents for hours or days without supervision

  • tackle large, loosely defined engineering projects across many repositories

  • need top results on research, quant finance or scientific analysis

  • have already tested Opus 5.5 at high effort and still need more

Stay on Opus 5.5 or Sonnet 5.5 if you:

  • do everyday coding, writing or analysis

  • care more about cost and speed than the last few points of quality

  • run high-volume calls where Fable's price adds up fast

How to access Claude Fable 5.2

  • Claude apps: Pro, Max, Ultra, Team and Enterprise plans.

  • Claude API: use the model ID claude-fable-5-2.

  • Cloud platforms: Amazon Bedrock (global.anthropic.claude-fable-5-2), Google Cloud, Microsoft Foundry and Claude Platform on AWS.

  • Claude Code: includes Fable 5.2 and its fast mode.

On effort, Anthropic's Fable 5.2 docs list five levels: low, medium, high, xhigh and max. The API default is high. Start there, since most of the gain is already present. Move to xhigh only for your hardest long-running work, where you've measured a real benefit. Apps and Claude Code default to medium. Read more in the effort guide.

Want to try Claude's coding agent on a budget? Here's how to use Claude Code free with AgentRouter.

Migrating from Fable 5.1 to Fable 5.2: a developer checklist

The model swap is one line:

model = "claude-fable-5-1"  # Before
model = "claude-fable-5-2"  # After

Then check these four things from Anthropic's what's new in Fable 5.2 page:

  1. Don't edit earlier turns. If your code changes the system prompt, tools or past messages, a request can fail with a 400 error. If Claude Code, claude.ai or the Agent SDK builds your requests, nothing changes for you.

  2. Watch model switching. Only Fable 5.2 and Mythos 5.2 can read Fable 5.2's thinking blocks. If a router or fallback hops to another model mid-conversation, that reasoning is dropped.

  3. Move computer use to the new toolset. The older computer_20251124 tool returns a 400 error on every platform except Amazon Bedrock. Use computer_toolset_20260801 instead.

  4. Update your advisor. A Fable 5.2 executor needs a Fable 5.2 or Mythos 5.2 advisor.

Two rules also apply to every request. Adaptive thinking is always on, so settings like thinking: {"type": "disabled"} fail. Forced tool use isn't supported either. Keep tool_choice on auto instead. Then rerun your evals with an effort sweep, and read the full migration guide.

Behavior changes to expect

Anthropic's prompting guide for Fable 5.2 flags a few habits:

  • Prompts like "answer directly without thinking" don't stop it from thinking at medium effort and above. Use a lower effort level for easy turns.

  • On long tasks, it sometimes ends a turn with a to-do list instead of finishing, most often at low and medium effort.

  • It follows style rules closely. A hard limit on sentence length can make its writing choppy.

  • Subagents can make up a large share of a session's cost, so set clear expectations for delegation.

Safeguards, Mythos and data privacy

Fable 5.2 blocks high-risk cyber and biology requests. A blocked request is answered by an older model instead: Opus 4.8 for cybersecurity and Opus 5 for biology. Anthropic explains how this works in its fallback help article.

If your legitimate work gets caught, you can apply to the Cyber Verification Program or the Life Sciences Verification Program for access to Mythos 5.2.

On privacy, Fable 5.2 requires 30-day data retention for safety monitoring. Zero data retention is only available if Anthropic expressly authorizes it. Enterprise Frontier Safeguards, now in private beta, aims to make monitoring work with zero data retention. Read the details in Anthropic's Enterprise Frontier Safeguards post.

How Fable 5.2 compares with GPT-6.2 Astra and Gemini 4 Neon

On BenchLM's index, Fable 5.2 scores 93.3, Neon 91.3 and GPT-6.2 Astra 90.8. Fable and Astra cost $8 / $40. Neon costs $3 / $15, which makes it the main rival if price matters. Anthropic's own table shows Fable 5.2 ahead of GPT-6 Astra and Gemini 4 Carbon on FrontierCode 2.0 (49.1% vs 42.8% and 41.1%) and FrontierSWE v2 (69.9% vs 65.5% and 61.2%).

For the whole field, see our top 20 frontier AI models ranking. To compare tools yourself, use our side-by-side comparison tool.

Should you upgrade from Fable 5.1?

Yes, for almost everyone. You get better results, faster output and a 20% lower price. The only cost is the migration check above. If your code is simple and doesn't edit history, the upgrade can take minutes.

The one reason to wait is a heavily tuned prompt stack. Fable 5.2 follows instructions more literally, so test your prompts before you move production traffic.

FAQs

What is Claude Fable 5.2?

It's Anthropic's newest and most capable generally available model for technical work. It launched November 9, 2026 with a 1M-token context window, 128K max output and a price of $8 / $40 per million tokens.

What's the difference between Fable 5.1 and Fable 5.2?

Fable 5.2 scores higher on every benchmark Anthropic reports, runs up to 30% faster, costs 20% less and blocks about 55% fewer requests. It also changes a few API behaviors, so check the migration notes.

How much does Claude Fable 5.2 cost?

$8 per million input tokens and $40 per million output tokens. Cache reads cost $0.20 per million, and batch pricing is $4 / $20. Fast mode costs $16 / $80.

Is Fable 5.2 better than Opus 5.5?

On most of Anthropic's benchmarks, yes. But Opus 5.5 costs half as much, and Anthropic recommends starting there for most coding and knowledge work. Move to Fable 5.2 for the hardest long-running tasks.

What is the difference between Fable and Mythos?

They're the same underlying model. Fable has strong cyber and biology safeguards and is open to all users. Mythos has looser safeguards and is only for vetted organizations in cybersecurity and life sciences.

What is the Claude Fable 5.2 API model ID?

Use claude-fable-5-2 on the Claude API, Google Cloud and Microsoft Foundry. On Amazon Bedrock, use global.anthropic.claude-fable-5-2.

Why was Claude Fable 5 suspended in June?

Anthropic suspended Fable 5 and Mythos 5 on June 12, 2026 to comply with US export controls. The controls were lifted June 30, and access was restored July 1.

Does Fable 5.2 support zero data retention?

Not by default. It needs 30-day retention for safety monitoring unless Anthropic expressly authorizes zero data retention. Enterprise Frontier Safeguards is designed to make the two work together.

Is Fable 5.1 still available?

Anthropic lists Fable 5.1 as a legacy model, and Fable 5 too. Fable 5's docs say retirement won't come before June 9, 2027. Check Anthropic's model deprecations page for exact dates.

Is Claude Fable 5.2 the best AI model right now?

On BenchLM's November 21 index, yes, at 93.3. The lead over Gemini 4 Neon (91.3) is small, and Neon costs far less. Check the live AI leaderboard for the latest scores.

Sources

Every spec, price and benchmark in this guide comes from Anthropic's own pages: the Claude Fable overview, the Fable 5.2 and Mythos 5.2 announcement, the Fable 5.2 system card, the Fable 5.1 and Mythos 5.1 announcement and the Fable 5 model docs. Independent scores come from BenchLM's Claude Fable 5.2 page and its Claude Fable 5.1 page. Some third-party model pages list outdated or wrong details, so we checked everything against Anthropic first.

Keep exploring

Want to see how Fable 5.2 ranks today? Open our AI leaderboard. Not sure which AI assistant fits you? Browse AI chatbots or take the AI tool matcher quiz. Building as a solo dev? See our picks of AI tools for developers and indie hackers.

Continue your research

Build a stronger shortlist

HyzenPro AI Tool Matcher

Want a faster path to the right AI tool?

Use the matcher hub to move from broad browsing into a guided shortlist based on workflow, budget, and team context.

About the Author

Rana Aqib
AI Workflow Researcher

Rana Aqib covers AI automation, video tools, coding assistants, and emerging productivity software. His reviews emphasize repeatable testing, buyer tradeoffs, and clear recommendations for different user types.

Expert Verified
Hands-on Testing

Share This Article

Related Articles