Anthropic announced on April 7, 2026, its newest frontier model, Mythos Preview which is part of the Claude family, as the most powerful it has developed. The company is withholding general release due to “unprecedented cybersecurity risks,” citing its superior ability to identify and exploit vulnerabilities in every major operating system and web browser.
Instead, it launched Project Glasswing, granting limited access to a consortium of 40 trusted partners including Amazon, Apple, Microsoft, Google and major banks for defensive vulnerability detection and patching. UK and US governments, including Treasury Secretary Scott Bessent, are actively engaged, with Anthropic labeled a prior “supply-chain risk” over Defense Department negotiations.
Trade analysis
Mythos is not publicly available, so current public leaderboards still feature Claude Opus 4.6 prominently alongside Gemini 3.1 Pro, yet internal data show Mythos leading 17 of 18 benchmarks: strong in coding, agentic tasks and more.
Bullish (YES) signals for Anthropic:
- Strong enterprise and government traction and prior agentic wins
- Sustained high Arena scores for Opus 4.6 family
Bullish (YES) signals for Google:
- Gemini 3.1 Pro leads public reasoning and multimodal benchmarks
- Broader accessibility and hardware integration
Markets price Anthropic as overwhelming favorite despite the restricted release, viewing Mythos as validation of superior underlying tech even if not vote-eligible. A strategy is to fade hype-driven spikes on restricted models or fear of non-public access, buying YES on Anthropic during pullbacks while remaining cautious on longshots like xAI. Our base case is Anthropic prevails through April via demonstrated model superiority and market momentum.