What is new
Anthropic has introduced Claude Fable 5.2 and Claude Mythos 5.2, two flagship models for the most demanding technical work. According to Anthropic, both share the same model foundation and differ mainly in access. Fable includes safeguards that block risky tasks in cybersecurity and biology and is available to all users. Mythos removes certain of these restrictions and is open only to organizations that apply through Anthropic's Verification Programs for Cybersecurity or Life Sciences. Anthropic says it deliberately prevented the models from gaining capabilities during training that could directly accelerate AI development.
Benchmark performance
The benchmark portal Artificial Analysis ranks Fable 5.2 first with 65 points on the Intelligence Index v5, five points ahead of Gemini 4 Carbon and 15 points ahead of its predecessor Fable 5.1. This figure refers to the highest compute tier. At the lower High and Xhigh tiers, Fable 5.2 still scores 62 points according to Artificial Analysis, ahead of every other model at every tier, with significantly lower token consumption.
In agentic knowledge work and terminal-based programming, Fable 5.2 according to the same sources sometimes trails competitors.
Price
Pricing is $8 per million input tokens and $40 per million output tokens, according to Artificial Analysis versus $10 and $50 for the predecessors. However, peak quality averages $26.50 per task according to Artificial Analysis, about 2.6 times the cost of Opus 5.5. Actual everyday costs depend on token consumption.
Cybersecurity
In an internal test called CVE Rediscovery, Mythos 5.2 found about 10.2 percent more vulnerabilities than Opus 5.5 and 14.4 percent more than Mythos 5.1 in the October 2026 run with 1,829 CVEs, according to Anthropic. A security firm reported to Anthropic that every sixth confirmed bug Mythos 5.2 found was missed by all other tested frontier models. The freely available Fable variant blocks most security-relevant tasks, such as penetration testing and writing attack code, according to Anthropic.
Cisco announced that it will add Mythos 5.2 to its AI Defense red-teaming offering, with authorization, attribution and audit capabilities, as Cisco writes.
Assessment
For German companies the access logic is the key point: according to the sources, using Mythos 5.2 requires a vetting process through Anthropic's Verification Programs. Whether and how this is open to particular use cases is not stated in the sources. The benchmark results come largely from Anthropic and Artificial Analysis and should be checked against one's own tasks before any concrete decision. Anthropic itself still recommends Claude Opus 5.5 for everyday use.
Sources
Editorially owned by Ideal Syka. Sources and method: Newsroom & method. Tips and corrections: ai@i6eal.de.
