Anthropic leak reveals new model "Claude Mythos" with "dramatically higher scores on tests" than any previous model

THE DECODER / 3/28/2026

📰 NewsSignals & Early TrendsIndustry & Market MovesModels & Research

共有:

Key Points

Leaked draft blog posts reportedly reveal Anthropic’s plans for a new model class positioned above its current Opus line, with “Claude Mythos” cited as a leading candidate name.
The documents claim the new model would achieve dramatically higher test scores than any previous Anthropic model.
Anthropic is said to be considering a deliberately slow release strategy, suggesting a cautious rollout rather than an immediate launch.
The leaked material indicates a strong emphasis on cybersecurity as a key focus area for the upcoming model class.

Update: The leaked draft blog posts have surfaced online, revealing Anthropic's plans for a new model class above its existing Opus line. The documents show two possible name candidates, details about a deliberately slow release strategy, and a strong focus on cybersecurity.

The article Anthropic leak reveals new model "Claude Mythos" with "dramatically higher scores on tests" than any previous model appeared first on The Decoder.

💡 Insights using this article

This article is featured in our daily AI news digest — key takeaways and action items at a glance.

📅 3/28DailyView insight →

Black Hat Asia

AI Business

Big Tech firms are accelerating AI investments and integration, while regulators and companies focus on safety and responsible adoption.

Dev.to

AI-Powered E-Commerce: Automating Product Descriptions at Scale

Dev.to

I shipped Google's TurboQuant as a vLLM plugin 72 hours after the paper — here's what nobody else tested

Dev.to

We built a governance layer for AI-assisted development (with runtime validation and real system)

Dev.to

Anthropic leak reveals new model "Claude Mythos" with "dramatically higher scores on tests" than any previous model

Key Points

💡 Insights using this article

Related Articles

Black Hat Asia

Big Tech firms are accelerating AI investments and integration, while regulators and companies focus on safety and responsible adoption.

AI-Powered E-Commerce: Automating Product Descriptions at Scale

I shipped Google's TurboQuant as a vLLM plugin 72 hours after the paper — here's what nobody else tested

We built a governance layer for AI-assisted development (with runtime validation and real system)

関連おすすめサービス

Notta搭載AI議事録イヤホン ZENCHORD1

AI搭載ボイスレコーダー Plaud

画像高画質化AIツール Aiarty Image Enhancer