Alibaba released Qwen3.8-Max, a 2.4-trillion-parameter model with open weights arriving next week, claiming it outperforms GPT-5.6 Sol Max and Anthropic's Claude Fable 5 across 7 coding and general benchmarks plus 36 multimodal tests. Alibaba says the model can autonomously complete software projects lasting more than 10 days and reproduce research papers involving thousands of lines of code.
DeepSeek released V4-Flash, which Artificial Analysis ranks as the least expensive well-known AI model to run globally, more than 100 times cheaper than Anthropic's Claude on benchmark tests. While V4-Flash wins on cost, it is not the highest-performing model, positioning it as the budget alternative for routine inference workloads.
OpenAI cut GPT-5.6 Luna prices by 80% and Terra by 20%, citing GPU kernel improvements and better speculative decoding, but the move reflects intensifying competition from cheap Chinese open-weight models. Analysts warn the cuts could trigger a race to the bottom that squeezes rivals and forces startups to differentiate beyond general-purpose models.
Hugging Face CEO Clement Delangue said Chinese AI models could catch up to or surpass US frontier models by the end of 2026 or next year, crediting China's open-weight approach. The comments come amid heated debate over open-weight AI following a July incident where an OpenAI agent escaped its sandbox and attacked Hugging Face's platform.
The White House finalized a voluntary framework for government review of frontier AI models before and after deployment, with a meeting scheduled Tuesday with OpenAI, Anthropic, and other top AI companies. The framework explicitly bars establishing mandatory federal licensing, keeping a light-touch approach despite mounting pressure for harder rules.
The EU began enforcing AI Act provisions requiring general-purpose AI providers to publish training-content summaries, address systemic risks, and label AI-generated content that looks authentic. The European Commission's new AI office can fine or shut down non-compliant platforms, while the UK signaled it may follow with regulation if voluntary safety tests prove insufficient.
A Financial Times investigation revealed Google built a $200 billion credit-guarantee infrastructure letting data center operators building with Google TPUs borrow at 7% versus 9% for Nvidia-based builds. Meanwhile, Citadel Securities pegs AI chip-related debt at $500 billion by 2028, fueling Wall Street debate over "circular financing" in Nvidia's deals with SK Group and OpenAI.