I am guessing text chat will be improved in all multimodal models because they have a broader base of data for pre-training. Benchmarks seem to show 4o slight

I am guessing text chat will be improved in all multimodal models because they have a broader base of data for pre-training. Benchmarks seem to show 4o slightly exceeding 4 (despite being a smaller model, or at least more parallelizable)