Google has officially released Gemini 2.5 Flash to all users, marking a significant advancement in both AI capabilities and security features.
The new version, now available in the Gemini app, brings major improvements across multiple dimensions. According to Google DeepMind head Demis Hassabis, Gemini 2.5 Flash is "better in nearly every dimension," showing enhanced performance on reasoning, multimodality, code, and long context benchmarks. Perhaps most impressively, it achieves these gains while becoming 22% more efficient, using fewer tokens to generate similar responses.
Security has been a major focus in this release. Google has implemented advanced safeguards that significantly increase Gemini's protection against indirect prompt injection attacks during tool use. These attacks occur when malicious instructions are embedded into data that an AI model retrieves. Through a combination of model hardening, input/output checks, and system-level guardrails, Google has made Gemini 2.5 its most secure model family to date.
Developers will gain access to the updated version in Google AI Studio and enterprises through Vertex AI in early June, with Gemini 2.5 Pro following shortly after. The release includes new developer-focused features like "thought summaries" that organize the model's reasoning process into a structured format with headers and key details, making interactions easier to understand and debug.
Gemini 2.5 Pro will be enhanced with Deep Think, an experimental reasoning mode specifically designed for highly complex math and coding tasks. This feature allows the model to consider multiple hypotheses before responding, leading to impressive scores on difficult benchmarks like the 2025 USAMO math test and LiveCodeBench for competitive coding.
The Gemini 2.5 family also introduces new audio capabilities, including native text-to-speech with support for multiple speakers across 24 languages. Users can customize the AI's tone, accent, and speaking style, with the model capable of detecting emotion in a user's voice and responding appropriately.
With these improvements, Google continues to position Gemini as a leading AI platform for both consumers and enterprises, balancing advanced capabilities with stronger security measures.