Data Machina #253
Substack 2 years ago 36
Google released multiple AI models and tools including Gemini 1.5 Pro with a 2 million token context window, Gemini 1.5 Flash optimized for efficiency, open-source models PaliGemma and Gemma 2, Project Astra interactive agents, and new safety and developer tools. Key benchmark: Gemini 1.5 Pro achieves over 99% retrieval accuracy up to 10 million tokens, exceeding Claude 3.0's 200k and GPT-4 Turbo's 128k context windows. Google's release of multiple capable models and open-source components shifts the landscape toward more accessible and specialized AI options across different use cases and organizations.