Google Launches Flash Cyber as Mythos Rival, Begins Pre-Training Gemini 4

Google is accelerating its AI model strategy on two fronts simultaneously, launching a new model called Flash Cyber that directly rivals Anthropic's Mythos series while also beginning pre-training on its next-generation flagship, Gemini 4.

Flash Cyber represents Google's latest attempt to balance AI capability with safety alignment. Anthropic's Mythos family has built a strong reputation for powerful reasoning capabilities combined with robust safety guardrails. By launching a direct competitor, Google signals it intends to compete aggressively in the high-performance, safety-tuned AI model segment.

In parallel, Google's next-generation flagship model Gemini 4 has entered pre-training. The Gemini series has been the cornerstone of Google's AI strategy, and this milestone suggests the company is laying groundwork for a major capability upgrade to its broader AI ecosystem.

The relationship between Flash Cyber and Gemini 4 remains unclear. Industry observers speculate Flash Cyber could be a standalone safety-enhanced product line, while Gemini 4 serves as the general-purpose flagship. Another possibility is that Flash Cyber's technical innovations will be integrated into the Gemini 4 architecture.

Google has been steadily increasing its AI investment in recent years. From the Gemini series to this new Mythos competitor, the company is engaged in an all-out model capability race with Anthropic, OpenAI, and others. The simultaneous launch of Flash Cyber and the start of Gemini 4 pre-training indicate Google is hedging its bets across multiple model product lines.

Mythos has established strong market credibility in AI safety. Whether Flash Cyber can find a better balance between safety and raw capability will be a key focus for the market.

With Gemini 4 entering pre-training, the industry is also watching for its release timeline and capability leap. Given that Gemini 3 already demonstrated significant advantages in multimodal and long-context capabilities, Gemini 4 is expected to push further in reasoning, code generation, and multimodal understanding.