DeepMind's SIMA 2 plays any game, MiniMax trains 2.7T model

Google DeepMind

Google DeepMind is shifting from AI that masters games to SIMA 2, a generalist agent powered by Gemini that understands and plays any game through natural language and visual input, partnering with game studios to create adaptive NPCs and transform game development.

Nvidia

NVIDIA's AVO agent architecture scored perfectly on the ARC-AGI-3 public benchmark using 12% fewer actions than comparable systems, demonstrating that long-horizon agent capability depends on system design (persistent memory, supervision, feedback loops) beyond just model quality. The same architecture previously achieved state-of-the-art results optimizing GPU kernels, proving its transferability across vastly different reasoning tasks.

Reuters

MiniMax is developing a massive 2.7 trillion parameter open-weight model targeting Q3 release, marking a major escalation in China's AI capability race and the broader competition for largest-scale models.

OpenAI

OpenAI is launching ChatGPT Ads across 31 European markets next week, its largest expansion yet, offering advertisers access to users during active decision-making moments while keeping paid tiers ad-free and maintaining privacy safeguards.

Nature Communications

Research shows diffusion model outputs typically cannot be attributed to specific training images, undermining copyright traceability arguments and complicating provenance claims for AI-generated content.