ICYMI: Google launched 3 experimental Gemini 1.5 models on AI Studio

Google has announced the release of three new experimental AI models: Gemini 1.5 Flash-8B, Gemini 1.5 Pro, and Gemini 1.5 Flash. These models are designed to gather feedback and provide developers with the latest updates, informing the broader release of the models in the future.

The New Feature: Gemini 1.5 Models

• Gemini 1.5 Flash-8B: A smaller, 8-billion-parameter variant of Gemini 1.5 Flash, targeting multimodal applications and long-context summarization. It is now freely accessible through Google AI Studio and the Gemini API under the name "gemini-1.5-flash-8b-exp-0827".
• Gemini 1.5 Pro: This model shows strong improvements across complex prompts and coding, serving as a drop-in replacement for the previous model. It is available via the Gemini API and Google AI Studio under the name "gemini-1.5-pro-exp-0827".
• Gemini 1.5 Flash: This model delivers significant performance improvements on many internal benchmarks. It is accessible via the Gemini API and Google AI Studio under the name "gemini-1.5-flash-exp-0827".

Key Capabilities

• Long Context Understanding: Gemini 1.5 can process and reason over vast amounts of information, with a context window of up to 10 million tokens, far exceeding other models like Claude 3.5 and GPT-4o.
• Multimodal Reasoning: The models seamlessly integrate and analyze information from text, images, video, and audio sources, enabling applications like visual question answering and content analysis.
• Function Calling: Gemini 1.5 can utilize external tools and APIs to perform more complex actions and tasks, opening up possibilities for building AI agents that interact with real-world systems and services.
• Instruction Following: The models accurately interpret and adhere to complex and nuanced instructions, ensuring responses align with user intentions and expectations.

Availability and Feedback

These experimental models are available for free via Google AI Studio and the Gemini API, with a free tier offered for both models. Google plans to release a version for production use in the coming weeks and encourages developers to provide feedback.