OpenAI released GPT-5.1 and GPT-5.1-Codex-Max, while other companies like xAI, Google DeepMind, and Anthropic introduced their own models. Notably, Anthropic's Claude Opus 4.5 is highlighted as the best option due to its unique approach with what they call a “soul document.” Instead of rigid behavior guidelines, Opus 4.5 emphasizes virtuous behavior and the reasoning behind its rules, leading to promising outcomes. In contrast, DeepSeek's v3.2 model performs well in benchmarks but suffers from speed issues and a lack of user enthusiasm. On the regulatory front, an attempt to bypass state AI regulations without a federal framework has failed for now. However, the article hints that this issue will resurface soon. The pace of significant developments has slowed following Thanksgiving, but several updates are still noteworthy. For instance, language models are demonstrating practical applications, such as solving science problems, while issues persist in accessing certain AI services, especially from Google. Various benchmarks have emerged, with Gemini 3 Pro leading in performance. Fun metrics like chess ratings and cybersecurity vulnerabilities illustrate the competitive landscape. Amazon introduced AI agents with limited details, raising concerns about their transparency. OpenAI is also preparing to introduce advertisements in ChatGPT responses, prompting mixed reactions from users who are worried about the implications for paid accounts.