It all makes sense. Google + Anthropic focused on pre-training, so you get to a great answer quicker, but the capability ceiling is lower. OpenAI focused on post-training, so great answers take longer, but you can push the models much further.

To be clear (getting a lot of confused comments), I obviously don't mean that any of these companies are solely focused on one thing. OAI/Anthropic/Google all focus on both pre-training and post-training (obv there are stages within those, just simplifying here). But it feels like OAI did a better job of post-training whereas Google/Anthropic did a better job of pre-training. They're all interlinked (better pre-training makes post-training easier/increases the ceiling for post-training, for example). This manifests in how the models feel to work with day-to-day.