AI News and information

Meta officially released the Llama 4 series multimodal open source model, adopting the MoE architecture, supporting graphic and text input, and has both efficient reasoning and multilingual processing capabilities.

Zapier launches MCP (Model Context Protocol) service, allowing AI assistants to connect more than 8,000 applications in one click without complex integration.

OpenRouter released the AI ​​model Quasar Alpha, which supports 1 million token context processing. It is now open for free and suitable for developers and content creators.

Google launches the new generation of AI model Gemini 2.5 Pro, which is now open for preview and supports developers to access and use through APIs.

OpenAI starts testing watermarks for images generated for GPT-4o, with restrictions only available to free users.

Quasar Alpha is released with a shocking release, supporting up to 1 million token context processing, and is optimized for code generation and complex inference tasks. Now available for free trial through OpenRouter.

The Browser Company launches the new AI browser Dia, which integrates GPT-4o and Gemini Flash 2.0, providing a personalized and concise browsing experience.

OpenAI released PaperBench evaluation benchmarks to evaluate the reproducibility of AI agents, and the Claude 3.5 Sonnet performed best.

Google NotebookLM has launched the "Discover Sources" function, which uses AI to automatically collect research data from the network, provide accurate recommendations, and greatly improve information sorting and research efficiency.

Tinder and OpenAI have launched the AI ​​voice flirting game "Game Game", allowing users to practice social interaction through virtual scenes.