OpenAI launches GPT-5.4-Cyber, challenging Anthropic’s Claude Mythos

OpenAI released GPT-5.4-Cyber, a cybersecurity-focused model, days after Anthropic unveiled Claude Mythos. The market for Google having the best AI model by end of June sits at 94% YES, while Anthropic’s odds for holding the third-best AI model by April have dropped under competitive pressure.

GPT-5.4-Cyber is tied to OpenAI’s Trusted Access for Cyber (TAC) program, which has expanded to thousands of users. Anthropic took a different approach with Claude Mythos, deploying it to a select group of partners. The Google AI dominance market for June 30 dipped slightly but remains at 94%.

Anthropic’s position has weakened. Maintaining the third-best AI model by end of April looks less likely now that GPT-5.4-Cyber directly challenges Claude Mythos’ positioning. Odds for Anthropic’s success in that market have dropped by an estimated 15% following OpenAI’s announcement.

The GPT-5.5 release market has 24-hour volume at $17,261 in actual USDC traded. The April 30 sub-market is priced at 91%, with $1,778 required to move odds by 5 points. The largest move was an 8-point spike, suggesting traders are hedging on near-term announcements. A YES share in the Google market currently pays out at 94¢, which leaves limited upside unless Google’s lead is seriously threatened.

OpenAI’s decision to scale cybersecurity access broadly rather than restrict it may not immediately affect Google’s AI ranking, but it introduces a variable that could shift positions before mid-year. Watch for new partnerships or model updates from either OpenAI or Anthropic that could move these markets before the April and June deadlines.

Get prediction market intelligence as a structured API feed. Early access waitlist.