Claude Opus 4.6 and GPT-5.3-Codex got released. They both claim to be better than all previous models, but obviously couldn't be compared to each other (yet).
OpenAI says GPT-5.3-Codex is the first model to classify as High capability for cybersecurity-related tasks under their Preparedness Framework. It's also meant to be 25% faster than its predecessor.
On the other hand, Anthropic claims they made substantial upgrades to Claude in Excel, and they're releasing Claude in PowerPoint in a research preview. The new model also features a 1M token context window in beta.
Who else is going to put these under the ringer today doing some tests? 😅