xAI Launches Grok 4.7. It's Bigger, But Late to the AI Frontier Party
Related
MIRA, an open-source 5B multiplayer world model, uses DINOv3 representation to mitigate long-term drift and can simulate 2v2 battles in Rocket League in real time.
According to Beating, AI research firm General Intuition, in collaboration with French AI lab Kyutai and Epic Games, has launched MIRA, a multiplayer interactive world model. As a generative game simulator supporting real-time multiplayer interaction, MIRA can simulate 2v2 battles in Rocket League in real time, based solely on historical footage and player button presses, without requiring a physics engine, rendering engine, or explicit 3D representation for inference. Unlike the "decoupling of logic computation and image rendering" approach adopted by companies like Odyssey, MIRA takes a generative simulation approach based on video latent space. MIRA boasts 5 billion parameters, and its core design builds the latent prediction space on a frozen general-purpose visual encoder, DINOv3-L. Leveraging pre-trained visual features, the generated latent states can more stably fall within the effective representation space, significantly mitigating image drift and divergence during long-term prediction. For multi-screen alignment, MIRA stitches the latent images from four player perspectives into a unified grid, enabling spatial attention mechanisms to operate naturally across viewpoints and improving the spatial consistency of vehicles, the soccer ball, and key events across multiple perspectives. The Action Dropout introduced during training also helps the system complete the game behavior of vehicles not controlled by commands when parts of the motion flow are missing. Currently, MIRA can run in real time at 20 frames per second on a single NVIDIA B200 graphics card. The team has open-sourced the training and inference code and released the Rocket Science dataset, which contains 1,000 hours of matches, approximately 4,000 hours of video, motion flow, and physics data from four perspectives; the complete training of the model used approximately 10,000 hours of clean match data.
GPT-5.6, Gemini 3.5 Pro, and Grok 4.5 will all be released soon.
PANews reported on July 7th that, according to market sources, Gemini 3.5 Pro will be officially released on July 17th. Its front-end and visual code generation capabilities are said to have seen a significant leap forward, outperforming Anthropic's Fable 5 in multiple tests. However, it still lags behind its competitor in hardcore inference and complex engineering tasks. Furthermore, Google DeepMind has abandoned the original 2.5 Pro platform, opting instead for completely new pre-training on Gemini 3.5 Pro, thus delaying the release date from the original June 2026 to July 17th. Elon Musk previously announced that xAI's latest large model, Grok 4.5, has begun beta testing within SpaceX and Tesla.
Nexo launches regulated crypto-backed credit in Australia
Nexo introduced regulated, cryptocurrency-backed credit lines that allow users to access liquidity in Australian dollars or stablecoins by borrowing against their digital assets.
Minnesota Says xAI's Grok Created 'Marketplace for Digital Sexual Violence'
Elon Musk’s AI firm says the state’s first-of-its-kind nudification ban violates the First Amendment. Minnesota says it regulates a tool, not speech.
The UK's FCA has warned of an "arms race" in the regulation of financial AI and is seeking new powers to regulate models such as ChatGPT.
PANews reported on July 6 that, according to the Financial Times, the UK Financial Conduct Authority (FCA) has warned of an "arms race" in the regulation of artificial intelligence in the financial sector and urged that it be given new powers to regulate ChatGPT, Claude, and Gemini.
Tether CEO warns AI giants' computing power subsidy model: Multiple cycles of mismatch continue to accumulate industry risks.
On July 4th, PANews reported that Tether CEO Paolo Ardoino published an article on the X platform, questioning the current expansion model of AI giants subsidizing computing power in exchange for user scale. He stated that leading global AI technology companies are continuously increasing their investment in computing infrastructure to seize market share, resulting in huge capital expenditures. However, the economic depreciation cycle of computing assets such as GPUs and servers is only 3 to 5 years, and the hardware depreciates extremely quickly. This creates structural mismatch risks: token prices are decoupled from the real value of assets, the profit realization cycle lags behind the capital investment cycle, and the cost of capital does not match the debt repayment period. At the same time, open-source AI models continue to divert market demand and compress commercial revenue space.