Claude Fable 5.2 and Mythos 5.2: What's New, Explained

On November 9, 2026, Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 — the same underlying model shipped in two safeguard tiers. Fable 5.2 is generally available today; Mythos 5.2 is the same model with lighter guardrails, limited to vetted organizations. If you only remember one thing: it's a step above Opus 5.5 on most evaluations, up to 30% faster than the 5.1 generation, and cheaper to run.

Here's what actually changed and whether it matters for how you build with agents.

Fable 5.2 and Mythos 5.2: one model, two tiers

Anthropic ships the pair as a single model behind two sets of safeguards:

• Fable 5.2 — the general-availability version with production safeguards. It's available today on all paid plans, on the Claude Platform, in Claude Code, and on all major cloud platforms.
• Mythos 5.2 — the same weights with more permissive guardrails, restricted to vetted organizations through Anthropic's Life Sciences and Cyber Verification Programs, for work in biomedical research and defensive cybersecurity.

In Anthropic's lineup, Fable sits at the top, above Haiku, Sonnet, and Opus. But Anthropic is explicit that it is not the model to reach for first: for most coding and knowledge tasks, it recommends starting with Opus 5.5, its strongest daily driver, then moving up to Fable or Mythos for the hardest, most open-ended technical work — or down to Sonnet 5.5 or Haiku 5.5 for lighter tasks.

What's new vs Fable 5.1

Three things stand out over the September 2026 Fable 5.1 release: price, speed, and raw capability.

1. Cheaper and faster

Fable 5.2 generates outputs up to 30% faster than Fable 5.1 and Mythos 5.1, and it costs less. Pricing drops to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 on the previous generation.

For teams running long, tool-heavy agent loops, cheaper-and-faster compounds: you pay less per token and finish sooner. Anthropic frames the pair as its strongest models for difficult technical work across coding, reasoning, and vision.

2. A step above Opus 5.5

Anthropic says Fable 5.2 and Mythos 5.2 are a step above Opus 5.5 in almost all the capabilities it evaluated, and that testers find they communicate more clearly than past Fable and Mythos models.

The coding claim comes with a cost angle. In early testing on an internal coding benchmark, Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort — for roughly half the cost per task. At the highest effort setting, Anthropic says it scored the highest of any released model it tested. Medium is the default in Claude apps and Claude Code, so most users won't hit those top-effort settings unless they dial them up.

Anthropic also points at its own codebase as evidence: it says over 90% of the changes merged into its core codebase are now written by Claude with human guidance, and that a team of agents running an early version of Fable 5.2 migrated a production service from Python to Rust, making it start about 26 times faster.

3. Beyond coding: robotics, math, and biology

The more striking demos are outside traditional software work:

• Robotics: Anthropic says Fable 5.2 shows substantially improved skills controlling physical robots, even on tasks it was never explicitly trained for. On RoboDojo, a benchmark with a simulated two-arm robot, it outperformed other models and most purpose-built robot-control systems.
• Formal math: a team of 1,000 Fable 5.2 agents fully formalized Fermat's Last Theorem in 30 hours — translating the proof into a form a computer can check step by step. With the same setup, Anthropic says Opus 5.5 completed only 26% in 18 hours and made little further progress after 30.
• Biology: Mythos 5.2 found a new way to predict a whole capsid's structure — the protein shell used to deliver gene therapies — accurately modeling 28 of the 34 capsids it attempted in Anthropic's tests.

These are company-reported results, not independent peer review. But they're concrete and point at where the frontier is heading: agentic systems doing open-ended technical work, not just chat.

Safety and external testing

Anthropic says Fable 5.2 and Mythos 5.2 performed as well as, or better than, its recent models on most alignment measures. The models were also tested before release by external evaluators, including the AI Security Institute and METR.

The split between the two tiers exists for safety reasons. Fable carries strong safeguards against misuse; Mythos lifts some of them for trusted organizations whose work — in cybersecurity or life sciences — needs fewer restrictions. That's why Mythos access runs through verification programs rather than a public API.

Should you switch?

A quick read by use case:

• Everyday coding and knowledge work — stick with Opus 5.5 first, per Anthropic's own guidance. Reach for Fable 5.2 when a task is genuinely hard or open-ended.
• Long-horizon, tool-heavy agents — Fable 5.2 is the pitch: faster output and lower token prices directly cut the cost of multi-step runs, and the Opus-beating coding scores suggest fewer failed attempts.
• Cost-sensitive teams — the $8/$40 pricing plus the "High effort ≈ Opus Max for half the cost" claim is worth a direct A/B test on your own workload before committing.
• Cybersecurity or life-sciences R&D — Mythos 5.2 is built for you, but only through the Cyber and Life Sciences Verification Programs.

The launch also lands as Anthropic readies for a widely expected IPO, a reminder of how central the Fable and Mythos line has become to the company's story this year. For a look at the prior generation and how the two-tier split started, see our Claude Fable 5.1 and Mythos 5.1 explainer.

For most builders, the honest summary is: a cheaper, faster top tier that's worth it specifically when the work is hard — not as your everyday default.

Do this with your own AI workforce

Fable 5.2's real edge is long-running, tool-heavy work — the same shape as the multi-agent tasks that fill a real workday. Eigent is an open-source "Cowork" desktop app that runs a multi-agent AI workforce locally, so you can point frontier models like Claude at coding, research, and document workflows on your own machine. See how it handles multi-step engineering in our review GitHub PRs workflow, or download Eigent and wire up your own agents today.