Claude Fable 5.2 (and Claude Mythos 5.2)

What it is

Claude Fable 5.2 is a large language model developed by Anthropic. The company released it on 9 November 2026 together with Claude Mythos 5.2, and says the two are the same underlying model ^[1]. What separates them is who can use each one. Fable 5.2 is available to all users and includes cybersecurity and biology safeguards ^[1]. Mythos 5.2 is available to vetted organizations through the Cyber Verification Program and the Life Sciences Verification Program, and to Claude Enterprise customers through Claude Security ^[1]. Fable 5.2 is available on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure, and its API model string is claude-fable-5-2 ^[1].

Background

Fable 5.2 succeeds Claude Fable 5.1 and Claude Mythos 5.1 ^[1]. Anthropic describes the new models as its most capable for demanding technical work, and says Fable 5.2 improves coding, reasoning and vision compared with Fable 5.1 ^[1]. The gains Anthropic highlights are in coding, finance and the sciences ^[1]. Most of the comparisons Anthropic publishes are made against Fable 5.1, the model it replaces, so that is the baseline used below ^[1].

Main features

Pricing is quoted per million tokens. A token is a unit of text that a model reads or writes, roughly a word or part of a word. Standard pricing for Fable 5.2 is $8 per million input tokens and $40 per million output tokens ^[1]. An optional fast mode costs $16 per million input tokens and $80 per million output tokens, and Anthropic says it speeds up token generation by up to 2.5 times ^[1]. Anthropic also says output is up to 30% faster than Claude Fable 5.1 ^[1].

Effort is the setting that controls how much work the model puts into each response. Medium effort is the default in Claude apps and Claude Code, and Anthropic says that on challenging tasks Fable 5.2 uses fewer tokens than Fable 5.1 at medium effort ^[1].

Anthropic's benchmark table compares the two models on several tests. On FrontierCode 2.0, Fable 5.2 scores 49.1% at Xhigh effort, against 32.4% for Fable 5.1 ^[1]. On Humanity's Last Exam with tools, the scores are 74.2% and 65.6% ^[1]. On OSWorld 2.1 (offline set, partial), Fable 5.2 scores 89.3%, against 83.5% ^[1]. On the Remote Labor Index, Fable 5.2 scores 41.3% at High effort, against 17.9% for Fable 5.1 ^[1]. These are numbers Anthropic published itself, and the effort level and test set matter when reading them ^[1].

Safeguards are the protections that stop a model from helping with harmful requests. Anthropic reports that safeguard blocks fell by about 55% overall, by about 60% in cybersecurity and by about 40% in biology, and says its cybersecurity and biology safeguards are as robust as those of Fable 5.1 ^[1]. Blocked queries fall back to Opus 4.8 for cybersecurity and to Opus 5 for biology, meaning another model handles them ^[1]. Distillation means training one model on the outputs of another. Anthropic says its anti-distillation protections reduced the data harvested by distillers by 90% to 99% ^[1].

Where it fits

Fable 5.2 is the latest step in the Fable series and the direct successor to Fable 5.1 ^[1]. Its companion Mythos 5.2 likewise succeeds Mythos 5.1 ^[1]. Released together, the two show one underlying model reaching different groups of users on different terms ^[1]. All the comparisons above are against Fable 5.1, so they measure the change from that predecessor rather than from any other model ^[1].