According to JeffDean, Google’s Gemma 4 12B is a unified multimodal model with open weights that runs on laptops under Apache 2.0.

According to StanfordAILab, a new YouTube-linked demo spotlights a Stanford AI Lab video understanding benchmark with metrics and research takeaways.

According to @gdb, OpenAI teased “fly with Codex” in a video, hinting at code-to-control demos. As reported by OpenAI’s post, developers await details.

According to @demishassabis, Gemma 4 tops 150M downloads and adds a 12B model that runs locally on 16GB VRAM under Apache 2.0 for laptop-grade multimodal AI.

According to Google Gemini on Twitter, a live demo shows multimodal inputs and conversational editing to create videos, streaming June 3 at 11:30am PT.

According to God of Prompt, Omni lags Seedance 2.0 in quality but wins with broader Google ecosystem integration and rapid deployment paths.

According to @gdb, GPT Realtime 2 enables full voice computer control, showcasing low latency, multimodal agents with OS actions in a live demo.

According to Ethan Mollick, Gemini Omni natively edits video via full multimodality, transforming the 1896 train film into multiple styled variants.

According to @godofprompt, SenseNova U1 unifies vision, language, and reasoning in one model, removing adapters and handoffs for higher fidelity.

According to GoogleDeepMind, Gemini 3.5 Flash has launched, signaling faster multimodal inference and lighter deployment for developers.

According to AndrewYNg, DeepLearning.AI launched a course on self-evaluating agents for image and video, combining similarity, LLM judges, and rubrics.

According to @godofprompt, Google I O 2026 reveals new Gemini models, personal agents, compute based pricing, and background web monitoring for operators.

According to @demishassabis, Google details Gemini 3.5 Flash’s fast multimodal performance and developer features on its official blog.

According to GoogleDeepMind, Gemini Omni enables multimodal story creation with text, images, and audio for faster prototyping and richer narratives.

According to KyeGomezB, ByteDance’s 3B Lance unifies vision tasks and outperforms 7B models via multi task synergy and MoE pathways.

According to DeepLearning.AI, a new poll challenges multimodal models to identify two gym machines, highlighting progress in visual reasoning.

According to OpenAI... ChatGPT users now create 1.5B images weekly, revealing fresh commercial design, prototyping, and marketing workflows.

According to DemisHassabis, Gemini Omni builds new scenes from photos, video, and audio, starting with video outputs and expanding to any input or output.

According to sundarpichai, Gemini 3.5 Flash shows strong evals as a workhorse model, signaling efficient multimodal performance for real-world apps.

According to @OriolVinyalsML, Gemini 3.5 Flash launches with frontier-level intelligence and faster speed, outperforming 3.1 Pro on most benchmarks.