Claude Fable 5.2 and Mythos 5.2: Anthropic's Updated Flagships for Complex Technical Work
Anthropic's Claude Fable 5.2 and Mythos 5.2 bring benchmark gains in agentic coding and research with lower pricing than prior versions; comparison of access.
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026. The two share the same underlying model but differ in safeguards and access. Fable 5.2 is generally available with strengthened cybersecurity and biology protections. Mythos 5.2 relaxes some of those restrictions and is limited to verified organizations through Anthropic's Life Sciences and Cyber Verification Programs.
Performance and Benchmarks
The new versions show gains across agentic coding, scientific research, and computer-use tasks. On FrontierCode 2.0, Fable 5.2 reached 49.1% at high effort compared with 32.4% for Fable 5.1. FrontierSWE v2 improved to 69.9% from 56.3%. In robotics simulation, the score rose to 64.9% from 27.6%. Visual chart recognition moved from 44.8% to 73.4%.
Against Claude Opus 5.5, Fable 5.2 leads on most agentic and multidisciplinary benchmarks while remaining competitive on knowledge work. Mythos 5.2 posts the highest scores on cybersecurity evaluations such as CyScenarioBench (76.3%) and CVE Rediscovery (89.6%).
Pricing and Speed
Input/output pricing is $8 per million tokens and $40 per million tokens, lower than the prior Fable 5.1 generation. An optional fast mode doubles the price for up to 2.5× faster generation. The models also produce output up to 30% faster than their predecessors at standard settings.
When to Choose Each Variant
Fable 5.2 suits ambitious software engineering, multi-step research, and agent coordination where strong safeguards are required. Mythos 5.2 targets organizations that need greater flexibility on cybersecurity or biomedical tasks and have passed verification. For everyday coding and knowledge work, Anthropic still recommends starting with Opus 5.5 and escalating only when scope demands it.
Limitations
Both models carry higher per-token cost than some competing offerings for routine use. Mythos access remains restricted. Even with improvements, the models are positioned for the hardest open-ended problems rather than as daily drivers for all users. Safety testing included novel evaluations for behaviors such as sandbox escapes, and the company reports that high-stakes misalignment risks remain low.
Independent benchmarks from prerelease testing and early user reports align with the official numbers on coding and agentic performance, though real-world cost per task depends on prompt complexity and chosen effort level.
Sources
Cover photo by Google DeepMind on Pexels, used under the Pexels License.
What supports the decision
Pros
- Leading scores on agentic coding and scientific benchmarks
- Lower pricing than Fable 5.1 generation
- Up to 30% faster output at standard settings
- Clear guidance on when to escalate from Opus 5.5
- Extensive safety testing with public system card
Cons
- Higher per-token cost than many daily-driver alternatives
- Mythos 5.2 access limited to verified organizations
- Positioned for hardest problems rather than general use
- Real-world performance varies with task complexity and effort settings
Key Specs
Was this worth your time?
One tap, no account. Your response helps decide what deserves a follow-up.
Comments (0)
Leave a Comment