Claude Fable 5.2: Specs, Pricing, Context Window and Availability

Claude Fable 5.2 is a proprietary large language model from Anthropic, released on 9 November 2026 for demanding reasoning and long-horizon agentic work. It is the highest-priced generally available Claude model, released alongside Claude Mythos 5.2, which has the same capabilities under restricted access. 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

• $8 input and $40 output per million tokens. Batch API requests cost half that, and optional fast mode costs $16 and $80 for up to 2.5x higher output tokens per second.
• 1M context and 128K output. The full 1M-token window is billed at the standard per-token rate, with no long-context surcharge.
• Thinking is always on. Adaptive thinking is the only mode and cannot be turned off; depth is set with the effort parameter, default high.
• 30-day retention is required. Fable 5.2 is a Covered Model on Anthropic's zero data retention page and is not available under ZDR unless Anthropic expressly authorises it.
• Some request patterns return 400 errors. Forced tool use (tool_choice any or tool), sampling parameters and edited history sent back with a later thinking block are rejected.

What is Claude Fable 5.2?

Claude Fable 5.2 is the third model in Anthropic's Fable line, after Fable 5 (9 June 2026) and Fable 5.1 (1 September 2026). It is one of five Claude model lines (Fable, Haiku, Mythos, Opus and Sonnet). The model page lists its intended uses as agent sessions that run for hours, multistep deep research, and analysis carried through to a finished document, spreadsheet or deck.

The announcement states that Fable and Mythos are the same underlying model. Fable carries cybersecurity and biology safeguards that make it available to all users, while Mythos 5.2 is limited to vetted organisations. The announcement also describes Fable 5.2 acting as an advisor to smaller models or as an orchestrator that delegates simpler tasks, and recommends starting with Opus 5.5 and moving up to Fable or Mythos, or down to Sonnet 5.5 or Haiku 5.5, depending on task scope.

Fable 5.2 is proprietary. Anthropic publishes no model weights and does not state a parameter count. For the full Claude line-up, see Anthropic Models; for the restricted sibling, see Claude Mythos (Preview, 5, 5.1, 5.2).

Key facts at a glance

As of 2026-11-15.

Release and version history

As of 2026-11-15. Dates are from the Claude Platform release notes and Anthropic's announcements.

Anthropic has not released a dated snapshot or point update of Fable 5.2 as of 15 November 2026. Fable 5 and Fable 5.1 remain Active.

How much does Claude Fable 5.2 cost?

As of 2026-11-15. USD per million tokens, from Anthropic's pricing page, the Fable 5.2 model page and the 9 November 2026 release note.

• Long context: the pricing page states that Claude 4.6 and later models, including Fable 5.2, bill the full 1M-token window at standard pricing.
• Fast mode: set speed: "fast" on a Messages API request, with no beta header. The pricing page lists fast mode on the Claude API, on Google Cloud (Anthropic managed and Google managed) and on Microsoft Foundry deployments hosted on Anthropic.
• Data residency: inference_geo: "us" applies a 1.1x multiplier to all token categories on the Claude API, Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed) and Microsoft Foundry US Data Zone deployments. Amazon Bedrock and Google-managed Google Cloud set their own regional pricing.
• Tokeniser: Fable 5.2 uses the tokeniser introduced with Claude 4.7 and later models, which the pricing page states produces roughly 30% more tokens for the same text than the previous tokeniser.
• Refusals: the Fable 5 release note states requests declined by a classifier before any output is generated are not billed; an opt-in fallbacks parameter re-runs refused requests on another model at that model's rates.

Compared with Fable 5.1 ($10 / $50, cache read $0.25), Fable 5.2's input, output and cache read prices are lower. See Model Pricing Explained and the LLM Pricing Tracker.

Context window and output limits

These limits match Fable 5.1. See Context Windows Explained.

Modalities and tools

• Input and output: text and image input, text output. The models overview states that all current Claude models support multilingual use, vision and tool use.
• Forced tool use: tool_choice types any and tool return a 400 error, as on Fable 5.1.
• Sampling parameters: omit temperature, top_p and top_k; setting them can return a 400 error.
• Prefix check: Fable 5.2 enforces the prefix check for every account. If code edits the system prompt, tools or an earlier message and then sends back a Fable 5.2 thinking block from a later turn, the request returns a 400 error.
• Thinking block portability: no other model reads Fable 5.2 or Mythos 5.2 thinking blocks; when a conversation moves to another model, including Opus 5.5 or Fable 5.1, the API drops them and the request still succeeds. Thinking blocks work only in the producing account or a linked account.
• Advisor tool: a Fable 5.2 executor accepts only Fable 5.2 or Mythos 5.2 as its advisor; any other advisor returns a 400 error.
• Dreams: the research preview supports Fable 5.2.
• Computer use and browser use: the toolsets were made available for the Fable line from Fable 5 on 19 August 2026; Fable 5.2-specific support is not separately stated on the pages checked (2026-11-15). See Computer Use and Browser Agents.

For how always-on reasoning affects cost and latency, see Reasoning Models Explained.

Where is Claude Fable 5.2 available?

As of 2026-11-15.

The model page links to Anthropic's supported countries page and does not list model-specific regions. In the Claude apps, the announcement states that eligible paid subscribers receive one to three Fable usage resets depending on tenure, and that Fable is included for up to 50% of weekly usage limits on Pro plans and standard Team seats.

Access restrictions

Fable 5.2 is generally available, with added safeguards that block certain tasks in high-risk, dual-use domains such as biology and cybersecurity. The announcement states that, compared with Fable 5.1, Fable 5.2 blocks about 55% less often (about 60% less in cyber and 40% less in biology), that blocked cyber queries fall back to Opus 4.8 and blocked biology queries fall back to Opus 5. The same capabilities without those Fable safeguards are offered as Claude Mythos 5.2, which is restricted to organisations verified through Anthropic's Cyber Verification Program or Life Sciences Verification Program, and to Claude Enterprise customers through Claude Security. Usage remains subject to Anthropic's usage policy and to the organisation's own API key and spend controls; see the Anthropic API Keys Security Guide.

Benchmarks (provider-reported, dated)

Scores as published by Anthropic in the Fable 5.2 announcement on 2026-11-09, using each model's highest available effort unless noted, and with safeguards on for Fable 5.2 and Opus 5.5. Only Claude models are shown, and rows reported only for Mythos 5.2 are omitted. No independent leaderboard figures are included on this page. See How to Read LLM Benchmarks.

The announcement also states that at High effort Fable 5.2 scores 45.3% on FrontierCode at roughly half Opus 5.5's cost per task.

Model card and system card

Anthropic published "System Card: Claude Fable 5.2 & Claude Mythos 5.2" on 9 November 2026, covering both models. What the card states:

• Deployment decision: "Claude Fable 5.2 is available for general use", while "Direct access to Claude Mythos 5.2 is limited to vetted individuals and organizations."
• Chemical and biological: treated as CB-1 with CB-1 protections; it "does not cross our CB-2 threshold."
• Autonomy: misalignment risk in high-stakes settings judged low, consistent with Anthropic's August 2026 Risk Report; the automated AI R&D threshold is judged not met, with the assessment described as uncertain.
• Cyber: the model "still falls within the lower category of capabilities" in Anthropic's Frontier Compliance Framework.
• Safeguards findings: Fable 5.2 falls back to Opus 4.8 more often under prompt injection, and regressed on multi-turn tracking and surveillance requests, which Anthropic states it is addressing with new classifiers.
• Alignment findings: the model appeared aware of possible evaluation in up to 43% of automated audit transcripts.
• External evaluators: the announcement names the UK AI Security Institute, Frontier Design and METR.

Anthropic does not publish a separate model card for Fable 5.2; the system card and the model page serve that purpose. See Model Cards and System Cards Explained.

See the Frontier AI Model Risk Frameworks Guide for how such reports are structured.

Data handling

• Retention: Anthropic's zero data retention page lists Fable 5.2 among the Covered Models that "require 30-day data retention. ZDR is therefore not available for any of them unless expressly authorized by Anthropic." Organisations with ZDR can enable 30-day retention for a single workspace other than the Default Workspace.
• Enterprise Frontier Safeguards: the announcement describes automated safety monitoring that runs on data stored in a customer's own infrastructure so it works with ZDR, in private beta; eligible customers can use the models with ZDR before adding it, and other customers have 30-day retention for safety monitoring.
• Training use: the ZDR page states "Retained data is never used for model training without your express permission."
• Flagged content: the ZDR page states Anthropic may retain inputs and outputs for up to 2 years where content is flagged.

See Zero Data Retention for AI, AI Data Retention and Deletion and Claude for Enterprise Security Guide.

Deprecation status

Anthropic defines Active as "fully supported and recommended for use" and gives at least 60 days' notice before retiring a publicly released model. The dates apply to Anthropic-operated platforms; Amazon Bedrock and Google-managed Google Cloud set their own schedules. See the LLM Deprecation and Retirement Tracker and AI Model Retirement and Decommissioning.

How Claude Fable 5.2 differs from other models in its family

As of 2026-11-15. Source: Fable 5.2 model page comparison table, models overview and release notes. Models listed by price, highest first.

• Against Fable 5.1: lower list prices, a published June 2026 knowledge cutoff, fast mode, the prefix check enforced for every account (Fable 5.1 enforces it by default only for accounts created on or after 31 August 2026), and a cheaper cache read ($0.20 against $0.25). The announcement states testers saw output up to 30% faster than 5.1. Anthropic publishes a migration guide from Fable 5.1 and from Opus models.
• Against Mythos 5.2: same capabilities, specifications and price; the difference is the safeguards and who may access the model.
• Against Opus 5.5: twice the list price, default effort high against medium, and thinking blocks that do not carry over if a conversation moves between the two.

Related profiles: Claude Opus 5.5, Claude Sonnet 5.5 and Claude Haiku 5.5.

Practitioner checklist

• Pin claude-fable-5-2 (or global.anthropic.claude-fable-5-2 on Bedrock) explicitly and record the platform, because lifecycle dates differ between Anthropic-operated and partner-operated platforms.
• Confirm the calling workspace has 30-day data retention enabled, and check that this matches your data protection commitments before routing regulated data to Fable 5.2.
• Remove temperature, top_p, top_k and forced tool_choice from requests, and test for 400 errors before switching traffic.
• Do not edit the system prompt, tools or earlier messages in a conversation that carries Fable 5.2 thinking blocks; start a new conversation instead.
• Decide whether to use the fallbacks parameter for classifier refusals, and budget for the fallback model's rates.
• Gate fast mode behind an explicit setting, since it doubles per-token prices, and apply spend limits; see How to set spend limits on LLM API keys.
• Measure output tokens at each effort level on real tasks, because thinking is always on and cannot be turned off.
• Read the system card's safeguards and prompt injection findings before giving Fable 5.2 agents tool or credential access; see the LLM Security Guide.

Standards and references

• Anthropic: Claude Fable 5.2 model page
• Anthropic: Introducing Claude Fable 5.2 and Claude Mythos 5.2
• Anthropic: System Card: Claude Fable 5.2 and Claude Mythos 5.2
• Anthropic: Models overview
• Anthropic: Pricing
• Anthropic: Claude Platform release notes
• Anthropic: Zero data retention
• Anthropic: Model deprecations

Related NHI Mgmt Group resources: Anthropic Models, Claude Mythos (Preview, 5, 5.1, 5.2), Zero Data Retention for AI, Anthropic API Keys Security Guide.