Part 11/13:
Both models support extended context lengths—up to 128,000 tokens—and include features like function calling, browsing, and structured reasoning. They follow instruction set constraints strictly; for example, if told never to say "five," they cleverly replace it with decimals to comply—showing how literal and rule-following these models are.
These models are ready to run on various platforms, including PyTorch on Apple Silicon, Windows (via ONNX), and cloud environments, with support from Microsoft and other industry players. Alongside, OpenAI launched a red-teaming challenge, offering $500,000 in prizes, encouraging safe deployment and thorough testing of open models.