engblogs

Gemini 2.5 Flash introduces a hybrid reasoning model with configurable thinking budgets, delivering enhanced accuracy, speed, and cost-efficiency for complex AI tasks via Google AI Studio and Vertex AI.

Exploring the cognitive processes and technical methods involved in interpreting and utilizing images.

Introducing OpenAI's latest models, o3 and o4-mini, showcasing advancements in AI capabilities and performance.

An overview and technical details of the OpenAI o3 and o4-mini system cards.

An overview of the enhancements made to the Preparedness Framework to improve efficiency and response capabilities.

OpenAI introduces a new nonprofit commission advisory board to guide its ethical and strategic decisions.

Veo 2 enables Gemini Advanced users and Whisk experiment participants to generate, animate, and share high-resolution AI-driven videos from text and image prompts with enhanced realism and safety features.

Announcing the release of GPT-4.1 in the API for enhanced AI capabilities.

Google's DolphinGemma AI model leverages advanced audio processing and machine learning to decode and interact with the complex vocalizations of Atlantic spotted dolphins, enabling new frontiers in interspecies communication.

GitHub introduces the sub-issues feature to enhance issue management by enabling hierarchical task breakdowns and improved project tracking through seamless integration and user-centric design.

Introducing StruQ and SecAlign, two fine-tuning defenses leveraging structured queries and preference optimization to robustly mitigate prompt injection attacks on Large Language Models without compromising utility.

Introducing BrowseComp, a benchmark designed to evaluate the performance of browsing agents.