OpenAI introduces GPT-5.4 with more knowledge work capability (arstechnica.com)

OpenAI has launched GPT-5.4, including the new GPT-5.4 Thinking and GPT-5.4 Pro models, as part of its expedited release strategy. This update is particularly notable as it arrives amidst growing competition from models by Anthropic and Google, aimed at retaining users who may consider alternative AI tools. The GPT-5.4 models are designed specifically for enhanced knowledge work, enabling tasks that require agentic inputs, such as issuing keyboard and mouse commands based on desktop or application screenshots. Key improvements in GPT-5.4 include more transparent reasoning capabilities, with the Thinking model providing upfront explanations and allowing users to adjust the AI's reasoning process dynamically. Additionally, the model boasts enhanced context maintenance for complex tasks and significant token efficiency, with a context window expanded to 1 million tokens. This makes GPT-5.4 a strong contender in the knowledge work domain, improving its utility for extensive web research and long-term tasks, and pushing the boundaries of what users can achieve with AI.