Claude Fable 5.2: Price, Access and What Mythos 5.2 Adds

What is Claude Fable 5.2?

Claude Fable 5.2 is Anthropic's upgraded generally available Fable model, launched on 9 November 2026 for "the most daunting technical challenges" such as large coding projects, research and engineering ^[1][6]. It follows Fable 5.1, which was released in early September ^[7].

Anthropic says Fable 5.2 improves on coding, reasoning and vision, that testers found it communicates more clearly, and that it generates output up to 30% faster than Fable 5.1 ^[1]. It can also act as an "Advisor" that smaller Claude models consult, or as an "Orchestrator" that hands simpler tasks to smaller models ^[1].

Anthropic's developer documentation lists Fable 5.2 with a 1 million token context window, up to 128,000 output tokens, always-on adaptive thinking, and training data up to June 2026 ^[2]. It supports text and image input and text output ^[2]. Anthropic says it will not retire the model sooner than 9 November 2027 ^[2].

The company does not recommend Fable for everyday use. It says most coding and knowledge tasks should start with Claude Opus 5.5, moving up to Fable only when needed ^[1][3].

How is Claude Mythos 5.2 different from Fable 5.2?

Mythos 5.2 and Fable 5.2 are the same underlying model; Fable adds strong cybersecurity and biology safeguards so it can be offered to everyone, while Mythos is reserved for vetted organisations ^[1].

Anthropic says Fable's classifiers block high-risk cyber work such as penetration testing, exploit generation and binary-based vulnerability research, as well as most research-biology tasks ^[1]. Blocked cyber queries are answered by Opus 4.8 instead, and blocked biology queries by Opus 5 ^[1].

Organisations working on defensive security or biomedical research can apply for Mythos 5.2 through Anthropic's Cyber Verification Program or Life Sciences Verification Program ^[1][3]. Claude Enterprise customers can also use Mythos 5.2 indirectly through Claude Security, a code-scanning product ^[1].

Anthropic first introduced Mythos in April 2026 with a limited rollout because of its cybersecurity abilities, and released Fable as the public, safeguarded version two months later, CNBC reported ^[3].

How much does Claude Fable 5.2 cost?

On the API, Fable 5.2 costs $8 per million input tokens and $40 per million output tokens ^[1][2]. CNBC reported that this is 20% less than the 5.1 generation ^[3], which Anthropic's documentation lists at $10 and $50 ^[2].

An optional fast mode generates tokens up to 2.5 times faster and costs $16 per million input tokens and $80 per million output tokens ^[1]. Fast mode is available on Claude Code, the Claude Platform, Claude Models on Google Cloud and Microsoft Foundry, and Mythos fast mode is limited to verification program members ^[1].

For comparison, Anthropic's documentation lists Opus 5.5 at $4/$20, Sonnet 5.5 at $2/$10 and Haiku 5.5 from $0.10/$0.50 per million tokens ^[2]. Fable 5.2 is therefore twice the price of Opus 5.5 per token, though Anthropic says that at High effort it can match Opus 5.5's best FrontierCode score for roughly half the cost per task ^[1].

Prices were checked on Anthropic's pages on 13 November 2026 and can change ^[2].

Can I use Claude Fable 5.2 on a Claude Pro plan?

Yes, but on Pro and Team standard seats Fable normally runs on paid usage credits; for launch week, Anthropic includes it in these plans until 16 November 2026 at 11:59 p.m. Pacific time ^[4][5].

• Pro and Team standard seats: during launch week, up to 50% of weekly usage limits can go to Fable models at no extra cost, plus one Fable limit reset that expires on 16 November ^[4].
• Max, Ultra, Team premium and legacy Enterprise premium seats: Fable is part of the plan, up to 50% of weekly limits, and the launch offer adds one to three resets depending on subscription length, valid until 8 March 2027 ^[4][5].
• Free plan: Fable models are not available ^[5].
• Enterprise plans: Fable 5.2 stays off until an admin enables it ^[5].

To use it, pick "Fable 5.2" in the model picker on web, desktop or mobile. Claude Code needs version 2.1.321 or later ^[5]. Resets only go to accounts with paid billing before 9 November 2026 at 10:00 a.m. PT ^[4].

Fable uses your limits faster than other Claude models, so check Settings > Usage while you try it ^[4].

How does Fable 5.2 compare with Opus 5.5, GPT-6 Astra and Gemini 4 Carbon?

On Anthropic's own benchmarks, Fable 5.2 beats Opus 5.5 on every score in the table below, and leads OpenAI's GPT-6 Astra and Google's Gemini 4 Carbon on most, but not all ^[1]. These numbers come from Anthropic and have not been independently verified.

On the FrontierCode 2.0 agentic coding test, Fable 5.2 scored 49.1% at its highest effort, against 45.5% for Opus 5.5, 42.8% for GPT-6 Astra and 41.1% for Gemini 4 Carbon (at medium effort) ^[1]. On FrontierSWE v2 it scored 69.9% ^[1]. Gemini 4 Carbon beat it on the AutomationBench business-workflow test, 53.9% to 51.4% ^[1]. On the AA-Briefcase knowledge-work rubric it scored 66.1%, against 61.2% for Opus 5.5 ^[1].

Anthropic also claims cost gains: at High effort, Fable 5.2 matched Opus 5.5's best FrontierCode score (45.3% vs 45.5%) for roughly half the cost per task ^[1]. Scores use each model's highest available effort unless marked, and some rival scores are missing ^[1].

What does Anthropic say Fable 5.2 has achieved?

Anthropic says that in prerelease testing Fable 5.2 solved two open problems in black-hole physics, formalized the proof of Fermat's Last Theorem in 30 hours, and designed three working electronic devices ^[1][6].

• Mathematics: about 1,000 Fable 5.2 agents formalized Fermat's Last Theorem in the Lean language in 30 hours, compared with 11 days for an earlier research version of Claude ^[1]. Anthropic amended this section on 11 November to describe more accurately the information it gave the model, which included a list of 60 statements and 47 proved lemmas from the earlier formalization ^[1].
• Software: one engineer had it coordinate over 1,000 subagents for eight days to build a prototype operating system that boots on a laptop ^[1].
• Hardware: it designed a voice-assistant hub, a USB speakerphone and a handheld virtual pet, which worked without electrical changes ^[1].
• Security: Mythos 5.2 found 10.2% more vulnerabilities than Opus 5.5 on Anthropic's internal CVE Rediscovery test ^[1].

These are company claims, some backed by preprints. Independent experts have not yet confirmed all of them, so treat them as demonstrations rather than guarantees of everyday performance.

How safe is Claude Fable 5.2?

Anthropic says it judges the risk of misalignment in high-stakes situations to be low, as with recent models, and that Mythos 5.2 scored better than recent Claude models in its automated behavioural audit ^[1]. External evaluators including the UK AI Security Institute and METR tested it before release ^[1].

In a deliberately weakened sandbox test with over 1,800 scenarios, Mythos 5.2 used credentials or files outside its sandbox in 3.1% of cases, similar to Opus 5.5 at 3.3% and far below Mythos 5.1 at 21% ^[1]. Anthropic also admits a limit: Fable 5.2 and Mythos 5.2 voice suspicion that they are being tested more often than recent models, which makes evaluations harder to trust ^[1].

For users, safeguards should block fewer legitimate requests: Anthropic observed about 55% fewer blocks than Fable 5.1 in early testing ^[1]. Eligible customers can use the models with zero data retention, and Enterprise Frontier Safeguards, which monitors prompts and responses stored in the customer's own infrastructure, is in private beta; other customers have 30-day retention for safety monitoring ^[1].

Should you switch to Claude Fable 5.2?

Only if your work is hard, long and technical; Anthropic itself recommends Opus 5.5 as the starting point for most tasks ^[1][2]. Anthropic's head of product management for research, Dianne Penn, told CNBC that Fable 5.2 is aimed at "power users or power use cases" ^[3].

• Good fit: large code migrations, multi-agent projects, research and engineering problems that Opus 5.5 cannot finish ^[1][2].
• Poor fit: quick questions, writing, and high-volume tasks, where Sonnet 5.5 or Haiku 5.5 cost far less ^[2].
• Security teams: Fable 5.2 blocks most offensive security work; apply for Mythos 5.2 instead ^[1].
• Pro subscribers: use launch week, until 16 November, to test it on your own tasks before paying with credits ^[4].

To see how Claude compares with rivals, read the Pouyam guide ChatGPT vs Gemini vs Claude in 2026: Models, Prices and Key Differences. For the basics of how these models work, see What Is ChatGPT? How Large Language Models Like It Actually Work.