Meta Platforms has released Muse Glimmer, a 30-billion-parameter AI model from its Superintelligence Labs that can run locally on consumer laptops and GPUs. The model is being distributed as open-weight software under an Apache 2.0 license, reigniting debate over whether powerful AI technology should be freely available.
French AI startup Mistral AI announced Shieldstral on August 4, 2026, a 3-billion parameter open-weights multimodal safety classifier that matches models up to seven times its size in performance while allowing operators to define custom moderation policies in natural language at runtime.
Hugging Face CEO Clément Delangue stated that China is winning the artificial intelligence race with open-weight models and could catch up to US model makers as soon as this year, following the recent OpenAI security breach affecting his platform.
The White House is excluding open-weight AI models from its voluntary framework to test advanced AI capabilities, even as Chinese developers dominate this category and industry leaders debate whether open models from China pose security risks or benefits. The policy shift comes amid growing concerns that China is winning the AI race through open-source releases.
Chinese AI company Moonshot AI has released the full weights of its Kimi K3 model, making the 2.8 trillion parameter system freely downloadable. The release raises questions about deployment practicality and trust in Chinese open-weight models despite the technical milestone.
The Trump administration is planning targeted bans on specific Chinese AI models rather than implementing blanket restrictions on open-weight models. The decision follows public pressure and an open letter from OpenAI and Google DeepMind opposing broad regulation.
Alibaba Group previewed its new Qwen3.8 artificial intelligence model, claiming it ranks second only to Anthropic's Claude Fable 5 in performance. The company plans to make the model open-weight soon, allowing developers to download and modify it.
Moonshot AI unveiled its Kimi K3 model, described as the world's largest open-weights AI system, which the company claims matches or exceeds OpenAI and Anthropic's leading models on some benchmarks. The release triggered market volatility in AI and semiconductor stocks as investors drew comparisons to last year's 'DeepSeek moment.'
Released July 15, 2026 by Thinking Machines Lab, Inkling is a 975-billion parameter open-weights Mixture-of-Experts model with 41 billion active parameters, a 1-million-token context window, and native multimodal input support across text, image, audio, and video.
AI startup Thinking Machines Lab, led by former OpenAI CTO Mira Murati, has released Inkling, a 975-billion parameter mixture-of-experts model with open weights designed for enterprise customization and on-premises deployment. The company explicitly positions the model as focused on customizability rather than benchmark dominance.
Chinese AI startup Z.ai (formerly Zhipu AI) has released GLM-5.2, a 753-billion parameter open-weight language model that achieved the highest score (51) among open models on the Artificial Analysis Intelligence Index and reportedly outperforms GPT-5.5 on long-horizon coding benchmarks at one-sixth the cost. The model features a 1-million token context window and is available under MIT license on Hugging Face.
Alibaba Group Holding has released its flagship AI model Qwen3.8-Max with 2.4 trillion parameters, making it accessible via QwenCloud. The company plans to release the model's weights for public download the week of August 10, with benchmarks showing performance approaching Anthropic's Claude Fable 5.
Meta Superintelligence Labs released Muse Spark, its first frontier AI model and first without open weights, marking a significant shift from Meta's open-source strategy. The model now powers Meta AI app and Meta.ai website.
Google's DeepMind released Gemma 4, a family of four open-weight AI models ranging from edge devices to workstations, marking the first time Google has used the fully open Apache 2.0 license instead of its previous restricted licensing. The models support up to 256K context windows and 140+ languages.
