How are token costs calculated?

Multiply input tokens by the input rate and output tokens by the output rate, add the two amounts and divide by 1,000,000. A standard request example uses 1,000 input and 500 output tokens. Monthly budgets multiply the result by request volume; taxes, hidden reasoning tokens, tools and tier surcharges can add costs.

Which recorded model has the lowest request cost?

LFM 2.2 6b by Liquid AI costs $0.00002 for the standard request, with $0.01 input and $0.02 output per million tokens. This price ranking does not establish quality equivalence; verify task success and context requirements.

Can cached and batch discounts be combined?

Only recorded rates are used. Cached input can have cache-write or storage charges, while batch rates apply only to eligible asynchronous jobs. We do not automatically stack the discounts or treat an absent rate as zero. Check the linked provider source before budgeting.

How current is the public dataset?

The live pricing table synchronizes stored model records daily; public file snapshots update when the site is rebuilt. Synchronization is not independent provider verification. It contains 1,028 priced models across 124 providers, last updated 2026-11-09. Directory entries are counted separately. A synchronization date is not proof of independent verification.

Analysis · 2026-11-09 · 11 min read

Claude Fable 5.2 & Mythos 5.2: $8/$40 Pricing, Benchmarks and Fast Mode Costs

Anthropic's Claude Fable 5.2 and Mythos 5.2 cost $8 per million input tokens and $40 per million output tokens, with a $16/$80 fast mode. Full pricing, benchmarks versus Opus 5.5, GPT-6 Astra and Gemini 4 Carbon, and when the premium is worth it.

Claude Fable 5.2 and Mythos 5.2 at a Glance

On November 9, 2026, Anthropic released Claude Fable 5.2 and Claude Mythos 5.2, its models for the hardest technical work. They are the same underlying model: Fable ships with cybersecurity and biology safeguards and is open to everyone, while Mythos removes those limits and is only available through Anthropic's Cyber and Life Sciences Verification Programs.

Anthropic says both models generate output up to 30% faster than the 5.1 versions and cost less.

Claude Fable 5.2 Pricing

The same prices apply to Mythos 5.2. Fast mode is available in Claude Code, the Claude Platform, Claude Models on Google Cloud and Microsoft Foundry (Anthropic hosted). Mythos fast mode is limited to Verification Program members.

Developers call it with the model ID `claude-fable-5-2`. Fable 5.2 is available on all platforms, including AWS, Google Cloud and Microsoft Azure.

How the Price Compares

Per token, Fable 5.2 is 20% cheaper than GPT-6 Astra and twice the price of Opus 5.5. Fast mode doubles the bill, so use it only where waiting costs more than tokens.

Benchmarks: Fable 5.2 vs Opus 5.5, GPT-6 Astra and Gemini 4 Carbon

Anthropic's launch table (safeguards on, highest available effort unless noted):

Mythos 5.2 scores 75.3% on Terminal-Bench-Science, 76.3% on CyScenarioBench and 89.6% on CVE Rediscovery. Gemini 4 Carbon still leads on AutomationBench.

The Cost Story: Fewer Tokens per Task

The headline price is higher than Opus 5.5, but Anthropic's cost-per-task charts tell a different story:

FrontierCode 2.0: at High effort Fable 5.2 scores 45.3%, matching Opus 5.5's best (45.5%) for about half the cost per task.
WANDR (agentic search): at Medium effort it matches Opus 5.5's top score for about 40% of the cost.
Humanity's Last Exam: at Low effort it matches Opus 5.5's top score for under one-tenth of the cost.
Chartography: 66.4% at Medium effort beats Opus 5.5 at Max effort for under one-tenth of the cost.

The lesson for budgets: a pricier model at a lower effort setting can beat a cheaper model pushed to maximum effort. Compare cost per finished task, not price per token.

What You Can Now Do with Fable 5.2 and Mythos 5.2

Anthropic's launch post describes what the models did during prerelease testing. These examples show the kind of work the higher price buys:

Build whole software projects with agent teams

Rewrite a production service in another language: agents running an earlier version of Fable 5.2 moved a critical Anthropic service from Python to Rust. It now starts about 26x faster and is more efficient and reliable.
Build an operating system from scratch: starting from an empty folder, Fable 5.2 coordinated over 1,000 subagents for eight days. The result boots on a laptop and has drivers, file storage, windows, a login screen and an installer.
Ship code your team approves: over 90% of changes merged into Anthropic's core codebase are now written by Claude, with human guidance. Testers at Vercel, Shopify and Datadog describe it as working like a staff-level engineer.

Find and fix security bugs (Mythos 5.2)

Mythos 5.2 is Anthropic's strongest model for cybersecurity engineering. It finds more vulnerabilities and is better at reproducing and fixing them.
One security company found that 1 in 6 confirmed bugs Mythos 5.2 caught were missed by every other frontier model they tested.
On CVE Rediscovery, which covers 1,829 real CVEs, it beat Mythos 5.1 and Opus 5.5 at every effort level.
Claude Enterprise customers can use it through Claude Security to scan codebases, even without joining the Cyber Verification Program. Fable 5.2 blocks most security tasks.

Do real science and math

Physics: Fable 5.2 solved two open problems about black holes using holographic duality.
Gene therapy: Mythos 5.2 built a method to model virus protein shells (capsids) used to deliver gene therapies. It modeled 28 of the 34 it attempted correctly.
Math proofs: a team of 1,000 Fable 5.2 agents formalized Fermat's Last Theorem in Lean in 30 hours, with minimal human help. An earlier research model took 11 days.

Design hardware and control robots

Electronics from a single paragraph: Fable 5.2 designed three working devices: a voice-assistant hub, a USB speakerphone and a handheld virtual pet. It wrote the specs, picked parts, laid out four-layer circuit boards, modeled 3D-printed cases and wrote the software. Each board was ready to manufacture in under a day, and all three worked without any electrical fixes.
Robots: it solved a puzzle cube, unscrewed a bottle cap and drew a picture. On RoboDojo Sim it scored 64.9%, more than double GPT-6 Astra (29.8%).

Research and analysis

Testers say it handles messier, data-heavy research with better judgment. Examples include quantitative research at Jump Trading, and finance and legal work at firms such as Balyasny, IMC, XTX Markets, Harvey and Box. One tester saw it flag mismatches between salary deposits and bank balances in a loan application.

What this means for cost: these are long-running jobs that use many agents. A 1,000-agent run uses a huge number of tokens at $40 per million output tokens. Budget for each job, set spend caps, and let cheaper models handle the routine steps.

When to Use Fable 5.2 (and When Not To)

Anthropic itself recommends starting most coding and knowledge work on Opus 5.5, then moving up to Fable or Mythos for the most ambitious, open-ended problems, or down to Sonnet 5.5 or Haiku 5.5 for simpler ones. Fable 5.2 can also act as an Advisor that cheaper Claude models consult only on hard steps, which keeps most tokens on lower-priced models.

Good fits: large code migrations, long-running multi-agent builds, research and hard reasoning. Poor fits: high-volume chat, classification and simple extraction, where Haiku 5.5 is 80x cheaper per token.

Key Takeaways

1. Price: $8 input / $40 output per 1M tokens; fast mode $16 / $80.

2. Same model, two doors: Fable for everyone, Mythos for verified cyber and life-sciences work.

3. Benchmarks: ahead of Opus 5.5 on nearly every listed eval and ahead of GPT-6 Astra on all shared ones.

4. Real cost: lower effort settings often reach Opus 5.5 quality for a fraction of the cost per task.

Check live prices on our pricing table and the Anthropic pricing page.