Multimodal

MiniMax released the open-source M3 model with coding, a 1M-token context window, and native multimodal support; M3 scores 59.0% on SWE-Bench Pro, 83.5% on BrowseComp, and costs about one-twelfth per token versus GPT-5.5.

Ethan He says a small xAI team built Grok Imagine from zero to one in 3 months, and the episode discusses video agents, audio-video alignment, inference speedups, and the storage, egress, and GPU-hour costs behind large video datasets.

NVIDIA open-sourced Cosmos 3 at GTC Taipei, releasing two variants, Super 32B and Nano 8B, with model weights, code, and datasets made available.

StepFun released Step 3.7 Flash, a sparse MoE model with 196B parameters plus a 1.8B ViT, activating 11B parameters per inference and reaching up to 400 tokens per second.

OpenAI has listed more than a dozen San Francisco robotics roles for OpenAI Robotics, a team that evolved from Aditya Ramesh’s Worldsim work, with the actuator design engineer role offering $342,000 to $445,000 in base cash pay plus PPU incentives.

VAST released Project Eden, a three-layer world-model architecture that separates persistent state evolution from visual rendering, and disclosed nearly $200 million across its A+ and A++ funding rounds.

CMG Lion Rock AI Lab released the LiOS edge-cloud architecture for embodied robotics, reporting about 30 ms one-way latency from local camera to cloud GPU memory in cross-machine tests, and open-sourced the low-latency video transmission module plus the LeFold laundry-folding dataset.

VAST raised nearly $200 million in A+ and A++ rounds and disclosed Project Eden, a world model architecture that separates state evolution from visual rendering through a structured state layer, a conditional interface layer, and a generative rendering layer.

Runway joined Cosmos Coalition as a founding member and will co-develop the first open world-model foundation model for physical AI with NVIDIA.

NVIDIA released Cosmos 3 as an open physical AI generalist model with native visual reasoning, world generation, and action generation, offering two variants: Super at 32B parameters and Nano at 8B parameters.

MiniMax released M3 as an open-source unified model with coding, agent, and native multimodal capabilities, supporting a 1M-token context window and using MiniMax Sparse Attention to cut per-token compute at 1M context to 1/20 of its predecessor, with over 9x faster prefill and over 15x faster decoding.

Qwen Studio lists seven capability areas: chatbots, image and video understanding, image generation, document processing, web search integration, tool use, and artifact generation; the post does not disclose Qwen3.7-Plus parameters, pricing, or release timing.

Vast raised nearly $200 million and reached a $1 billion valuation; the RSS snippet says the 3D-modeling startup was founded by a 29-year-old gamer but does not disclose investors or product specifications.

OpenAI's June 2026 threat report details a banned cluster of ChatGPT accounts likely originating in China. The operators used Simplified Chinese prompts and VPNs to generate English comments and political cartoons criticizing US tariffs, rare earths, AI, and 5G policy. They instructed the model to depict only Trump, not Xi Jinping or China. The same cluster produced Chinese-language water army content attacking the US and Israel, amplifying anti-Jewish tropes, and harassing dissidents. OpenAI also linked these accounts to a separate X network that falsely claimed ChatGPT user data was compromised. The post does not disclose the exact number of banned accounts or the operators' specific institutional affiliation.

MouShen Intelligence released STI-WM, a spatiotemporally integrated world-action model for robotics, claiming support for RGB, point-cloud, and proprioceptive inputs, hundred-second task planning, and disclosing five funding rounds in six months plus a RMB 300 million Pre-A round.

Moushen Intelligence released STI-WM, a spatiotemporally integrated world action model for robotics, with RGB, depth point cloud, and proprioceptive inputs; the post says it supports hundred-second-scale long-horizon task rollout and closed-loop replanning, but does not disclose benchmark scores or deployment costs.

Sellers use AI-generated Black personas on TikTok, Facebook, and Instagram to pose as handmade creators and sell mass-produced dropshipped goods; the snippet cites one fake persona, “Aliyah,” selling fictional handmade belt buckles.

Google AI Developers released Nano Banana Pro and Nano Banana 2, mapped to gemini-3-pro-image and gemini-3.1-flash-image. The post says both are production-ready through the Gemini API, but does not disclose pricing, benchmarks, or runtime limits.

Yi Tay served as a modeling co-captain for Gemini Deep Think when it reached IMO gold-medal level, co-founded Reka AI in 2023, and returned to Google DeepMind after 639 days, while the article also notes his 2012 Trinity classical piano associate diploma.

Apple’s team published PICO, a perceptual image codec that uses 57%-70% fewer bits than AV1, VVC, and JPEG AI at the same subjective visual quality, while encoding a 12MP photo in 230 ms and decoding it in 150 ms on an iPhone 17 Pro Max.