Claude Fable 5.2 vs Fable 5.1: What Changed, Specs, Pricing and Availability

Claude Fable 5.2 and Claude Fable 5.1 are two proprietary Anthropic models in the Fable line, released on 9 November 2026 and 1 September 2026, and both remain callable on the Claude API. This page compares the two versions field by field: API IDs on each platform, status, pricing, context and output limits, documented behaviour changes, breaking changes for code, availability, system card findings, provider-reported benchmarks and retirement dates. It is written for developers deciding whether and how to move from claude-fable-5-1 to claude-fable-5-2, and for buyers checking cost. Versions are listed newest first, an objective order that implies no ranking. Every fact comes from Anthropic's own documentation, announcement and system cards, cited in the sources list.

Last checked: 2026-11-15.

This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It uses Anthropic's public documentation as of November 2026. Model details change, so confirm against Anthropic's current documentation.

Key takeaways

• Fable 5.2 lists 20% lower prices. Anthropic's pricing page shows $8 input and $40 output per million tokens for Fable 5.2, against $10 and $50 for Fable 5.1, with cache and batch prices lower by the same proportion.
• Limits are unchanged. Both models list a 1M-token context window, 128K maximum output, text and image input, always-on adaptive thinking, default effort high and a June 2026 knowledge cutoff.
• Anthropic lists four breaking changes. The migration guide names prefix check enforcement, model-bound thinking blocks, rejection of computer_20251124 outside Amazon Bedrock and a stricter advisor pairing.
• Fast mode is Fable 5.2 only. The pricing page lists fast mode at $16 and $80 per million tokens for Fable 5.2 and lists no fast mode price for Fable 5.1.
• Both are Active. The deprecations page lists Fable 5.1 as Active with retirement not sooner than 1 September 2027, while its model page labels it legacy and recommends migrating to Fable 5.2.

What is Claude Fable?

Fable is one of five Claude model lines in Anthropic's documentation, alongside Haiku, Mythos, Opus and Sonnet. The line began with Claude Fable 5 on 9 June 2026, followed by Fable 5.1 on 1 September 2026 and Fable 5.2 on 9 November 2026. Each Fable release has shipped with a Mythos sibling: Anthropic's Fable 5.2 announcement states that Fable and Mythos are the same underlying model, with Fable carrying cybersecurity and biology safeguards that make it available to all users and Mythos limited to vetted organisations.

Anthropic's models overview recommends Fable 5.2 "for demanding reasoning and long-horizon agentic work, or when your evals on Claude Opus 5.5 at higher effort still fall short." The overview lists Fable 5.2 under its latest models and Fable 5.1 only in its all models table. Both are proprietary: Anthropic publishes no weights and states no parameter count. For the wider line-up, see the Anthropic Models hub; for the full single-model profile, see Claude Fable 5.2; for the restricted sibling line, see Claude Mythos Versions Compared.

Every model at a glance

As of 2026-11-15. Source: Anthropic models overview, model pages and model deprecations page.

Both models take text and image input and return text. Open weights: none for either version. Licence: proprietary. Parameters: not stated by Anthropic. On Microsoft Foundry, Claude Platform on AWS and both Claude Models on Google Cloud options, the ID matches the Claude API ID for each model, per the model pages.

When was each model released?

As of 2026-11-15. Dates are as shown in Anthropic's Claude Platform release notes.

How much does each model cost?

As of 2026-11-15. USD per million tokens, from Anthropic's pricing page and the two model pages.

Cache read ratio: the Fable 5.1 release note states its cache read is 0.025x base input, against 0.1x on other models; Fable 5.2 keeps the same ratio ($0.20 is 0.025x of $8).

Long context: the pricing page states that Claude 4.6 and later models, including Fable 5.1 and 5.2, bill the full 1M-token window at standard pricing, and that caching and batch discounts apply across the full window.

Tokeniser: both versions use the tokeniser introduced with Claude Opus 4.7, which the pricing page states produces about 30% more tokens for the same text than the tokeniser used by Sonnet 4.6 and earlier. Prices per token between the two Fable versions are therefore directly comparable with each other.

Data residency: for Claude 4.6 and later models, setting inference_geo to "us" applies a 1.1x multiplier to all token categories. For cost guardrails, see How to set spend limits on LLM API keys and the LLM Pricing Tracker.

Worked cost example

Using the list prices above, a workload of 1,000 requests a day, each with a 50K-token prompt and a 4K-token response, uses 50 million input and 4 million output tokens a day. On Fable 5.1 that is $500 input plus $200 output, $700 a day. On Fable 5.2 it is $400 input plus $160 output, $560 a day. If 80% of each prompt is a cache hit, input on Fable 5.2 becomes $80 for the uncached 10 million tokens plus $8 for 40 million cached tokens, before cache write charges. Thinking tokens count as output and vary with effort, so measure output on your own tasks; Anthropic's announcement states that at Medium effort Fable 5.2 completes tasks with fewer tokens than 5.1.

What changed between models?

Only changes Anthropic documents in its release notes, migration guide, What's new page and announcement are listed. The migration guide describes the move as "mostly drop-in" and states that the API surface, always-on adaptive thinking, the forced tool choice restriction, refusal handling, stop_details categories and the data retention requirement match Fable 5.1.

Breaking changes listed by Anthropic

1. Prefix check enforced for every account: if code changes system, tools or an earlier message and then sends back a Fable 5.2 thinking block, the request returns a 400 error. On Fable 5.1 the error applies by default only to accounts created on or after 31 August 2026. The guide states Claude Code, claude.ai, Managed Agents and the Agent SDK manage history automatically, and that an MCP connector without a pinned tool list also needs testing, because a changing tool list counts as an edit.
2. Thinking blocks are model-bound: Fable 5.2 reads Fable 5.1 thinking blocks on the Claude API and Google Cloud (Google managed), but Fable 5.1 cannot read Fable 5.2 blocks. If a router, retry or refusal fallback moves a conversation from 5.2 to 5.1, the API drops the unreadable blocks; the request succeeds and the dropped input tokens are not billed, but the next turn may cost more and run slower.
3. computer_20251124 rejected: it returns a 400 error on every platform except Amazon Bedrock; declare computer_toolset_20260801 instead.
4. Advisor pairing: under the advisor tool (beta), a Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor; Fable 5.1 as advisor returns a 400 invalid_request_error. A Fable 5.2 advisor returns its advice as an encrypted advisor_redacted_result block.

Behaviour changes listed by Anthropic

• "Think less" prompt lines: a line such as "answer directly without deliberating" usually works on Fable 5.1, but in Anthropic's testing did not stop Fable 5.2 thinking at medium effort and above. Anthropic advises lowering effort instead.
• Subagent self-review: in agentic coding at xhigh and max effort, Fable 5.2 was observed starting subagents to review its own work, which Fable 5.1 did not do on the same tasks.
• Small max_tokens: thinking counts toward max_tokens, so a response can stop after a thinking block and before any text.
• Rate limits: Fable 5.2 has its own rate limit, separate from the combined limit shared by Fable 5.1 and Fable 5.
• Safeguard blocks: the announcement states early testing across claude.ai, Claude Code and Cowork showed about 55% fewer safety classifier blocks than 5.1, about 60% fewer in cybersecurity and 40% fewer in biology.
• Speed: the announcement states testers saw outputs up to 30% faster than 5.1.

For background on effort and adaptive thinking, see Reasoning Models Explained.

What workloads does Anthropic position each model for?

This section records the provider's own positioning wording; it is not an assessment.

What is the context window of each model?

As of 2026-11-15. From the two model pages and the release notes.

On both versions, thinking cannot be disabled and manual thinking budgets return a 400 error, as introduced with Fable 5. Fable 5.2's What's new page lists the effort levels and advises xhigh for the hardest long-running coding and agentic work and max for measuring the ceiling. See Context Windows Explained.

Capabilities and tools compared

As of 2026-11-15. From the release notes, the Fable 5.2 What's new page and the migration guide.

The Fable 5.1 release note states its text carries Anthropic's text watermark and that files produced by the code execution tool carry C2PA Content Credentials when retrieved through the Files API. A watermark statement specific to Fable 5.2 is not made on the pages checked (2026-11-15). See the Computer Use and Browser Agents guide.

Where is each model available?

The Fable 5.1 release note lists Google Cloud (Google managed) without the Anthropic managed option; the current model page lists both. Partner-operated platforms publish their own model catalogues and schedules, so confirm in each one.

Model cards and system cards

What each card states, in summary:

• Deployment: both cards describe the Fable model as the general-use release with additional safeguards that block certain high-risk biology and cybersecurity tasks, and the Mythos sibling as limited to vetted individuals and organisations.
• Chemical and biological: both cards treat the model as CB-1 and state it does not cross the CB-2 threshold.
• Cyber: the Fable 5.1 card places Mythos 5.1 in Tier 1 of Anthropic's Frontier Compliance Framework, "getting closer to Tier 2", and states Fable 5.1 allows vulnerability discovery in source code but blocks it in compiled binaries. The Fable 5.2 card states the model still falls in the lower capability tier.
• Fallbacks: the Fable 5.1 card states flagged requests fall back to Opus 4.8 on most interfaces; the Fable 5.2 card states biology-blocked requests go to Opus 5 and cyber-blocked requests to Opus 4.8, with opt-in fallbacks on the API.
• Fable 5.2 against 5.1: the Fable 5.2 card describes safeguards as largely comparable, with improved multi-turn biology handling, a lower single-turn harmless response rate and a regression on multi-turn tracking and surveillance requests, which Anthropic states it addressed with new classifiers. Under prompt injection, Fable 5.2 falls back to Opus 4.8 more often.

See Model Cards and System Cards Explained.

Benchmarks

As of 2026-11-15. Figures are provider-reported by Anthropic on 2026-11-09, in the Fable 5.2 announcement or system card, generally at each model's highest available effort with safeguards on unless noted. Rows reported only for the Mythos variants are omitted. No independent leaderboard figures are included. See How to Read LLM Benchmarks.

The Fable 5.1 system card does not report Fable 5.1 cyber scores, stating that its classifiers consistently fire during testing.

Deprecations and retirements

As of 2026-11-15. Source: Anthropic model deprecations page. Anthropic defines Active as "fully supported and recommended for use" and gives at least 60 days' notice before retiring publicly released models.

The dates apply to Anthropic-operated platforms (Claude API, Claude Platform on AWS, Claude Models on Google Cloud Anthropic managed, Microsoft Foundry); Amazon Bedrock and Google-managed Google Cloud set their own schedules. See the LLM Deprecation and Retirement Tracker.

Data handling and security

Anthropic's zero data retention page designates Fable 5.2, Fable 5.1 and their Mythos siblings as Covered Models that "require 30-day data retention. ZDR is therefore not available for any of them unless expressly authorized by Anthropic." Organisations with a ZDR arrangement can enable 30-day retention for a single workspace other than the Default Workspace. The page states retained data is never used for model training without express permission, and that flagged chats or sessions may be retained for up to 2 years. The Fable 5.2 announcement describes Enterprise Frontier Safeguards, safety monitoring on data held in the customer's own infrastructure, in private beta.

The migration adds one access-control detail: Fable 5.2 thinking blocks work only in the producing account or a linked account, so stored conversations must be replayed through that account. Claude API keys reach every model a workspace can call; see the Anthropic API Keys Security Guide and Zero Data Retention for AI.

How to choose a model: factual criteria

Run your own effort sweep and evaluation on representative tasks before switching. For side-by-side line comparisons, see Claude Fable vs Mythos vs Opus and Frontier AI Models Compared.

Practitioner checklist

• Inventory every service and agent calling claude-fable-5-1 or anthropic.claude-fable-5-1 and record the platform each runs on.
• Check the separate Fable 5.2 rate limit before moving production traffic.
• If your code builds messages itself, run the prefix check test with the thinking-binding-controls-2026-08-01 header and fix every prefix_binding_mismatch entry.
• Pin MCP connector tool lists, because a changing tool list counts as an edit for Fable 5.2 thinking blocks.
• Replace computer_20251124 with computer_toolset_20260801 outside Amazon Bedrock and update the agent loop.
• Set the advisor tool model to claude-fable-5-2 or claude-mythos-5-2 for Fable 5.2 executors, in new conversations.
• Remove "think less" prompt lines, set effort explicitly and raise max_tokens where thinking could leave no room for text.
• Gate fast mode behind an explicit setting and set workspace spend limits.

Standards and references

• Anthropic: Migrating to Claude Fable 5.2 (accessed 2026-11-15)
• Anthropic: What's new in Claude Fable 5.2 (accessed 2026-11-15)
• Anthropic: Claude Fable 5.2 model page (accessed 2026-11-15)
• Anthropic: Claude Fable 5.1 model page (accessed 2026-11-15)
• Anthropic: Pricing (accessed 2026-11-15)
• Anthropic: Claude Platform release notes (accessed 2026-11-15)
• Anthropic: Model deprecations (accessed 2026-11-15)
• Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2 (accessed 2026-11-15)
• Anthropic: System Card: Claude Fable 5.2 and Claude Mythos 5.2 (accessed 2026-11-15)
• Anthropic: System Card: Claude Fable 5.1 and Claude Mythos 5.1 (accessed 2026-11-15)
• Anthropic: Zero data retention (accessed 2026-11-15)

Related NHI Mgmt Group resources: Anthropic Models, Claude Fable 5.2, Claude Fable vs Mythos vs Opus, LLM Pricing Tracker, Claude for Enterprise Security Guide.