The models are available in three sizes: Flash (small), Pro (medium), and Ultra (large), catering to a range of user needs.

Investor scrutiny has intensified due to rising costs associated with AI model development, particularly following DeepSeek's reported training costs of under $6 million.

Google has launched Gemini 2.0 Flash, making its AI models accessible through the Gemini API in Google AI Studio and Vertex AI.

The new Gemini 2.0 Pro model boasts superior coding performance and can manage complex prompts with a 2 million token context window, ideal for analyzing large datasets.

In response to positive feedback, Gemini 2.0 Flash-Lite has been developed to enhance quality while maintaining speed and cost efficiency.

Users can access the new models for free, except for the subscription-required Pro model, as Google continues to address safety risks through red teaming.

The emergence of cheaper AI alternatives has prompted major tech companies, including Alphabet, Microsoft, and Meta, to reaffirm their commitment to substantial investments in AI.

All models in the Gemini suite prioritize safety and responsibility, employing techniques like reinforcement learning and automated red teaming to ensure ethical use.

Safety is emphasized across the Gemini 2.0 family, with reinforcement learning implemented to improve response accuracy and manage sensitive prompts.

This launch marks a significant shift as Google's models are now integrated into external services like Perplexity, indicating a move towards greater openness with AI startups.

The model's ability to trace its reasoning represents a significant advancement in AI explainability, enhancing user trust.

The new models are accessible via the Gemini app, web interface, Google AI Studio, and Vertex AI, although they remain experimental.