Developers promise a context window of 2 million tokens, which is 16 times more than GPT-4o.
GPT-4o has 128k tokens, equivalent to two copies of "The Great Gatsby" while Gemini 1.5 Pro can load the entire "War and Peace".
Gemini will now be integrated into ALL Google products: Search, Gmail, Photos, Workspace, NotebookGmail, and Google Meet. The model can write emails, summarize them, engage in dialogue, search for relevant parts of emails, read attachments, and respond to any lengthy documents, videos, or images in attachments. It can be controlled by voice.
Google also introduced Gemini 1.5 Flash, an optimized model with low latency.
The new Project Astra is a prototype from Google DeepMind featuring AI assistants that can communicate in real-time. The AI operates directly from your phone and even smart glasses! Project Astra could be genuinely useful in everyday life.
The agents can interact with the surrounding world, perceive information, remember what they see, process this information, and understand the environment and details.
Veo has been introduced as a direct competitor to Sora for video generation. The model accepts text and can generate videos up to 1080p resolution lasting over a minute.
Imagen 3 has been unveiled as Google's most advanced model for image generation.
Google is finally making serious efforts to integrate artificial intelligence into its search engine.