Model Wars: Hype or real?

Don’t waste time arguing over which model is better, because they are all good enough.

Across all the model providers (OpenAI, Anthropic, xAI, etc) the performance difference in the latest models is marginal.

Given the right prompt, any of the latest generation models can generate the text needed for most use cases.

My personal approach: Just use OpenAI’s latest models.

If the AI can’t complete your task, it is more likely due to poor prompting than it is to choosing the wrong model.

Here’s how to improve your prompt step by step:

1. Format: Ensure you’ve described what you’re looking for — “less than 10 words”, “JSON format”.
2. Context: Ensure the AI has all the relevant background info needed (e.g. user bio, last action taken).
3. Examples: Add one example, then a few more if you’re still struggling.

Reach out to us for more advice on prompt engineering.

I’ve found that the best AI engineers don’t argue over what model is being used. It is almost always the gurus or those who are just getting started in AI. This is likely because those people don’t know how to prompt the models. Great artists can still produce great works with cheap tools.

Often, pricing, latency, tool calling, or developer usability matter more than another point or two higher intelligence score.

This is much like when Apple stopped promoting their computers by boasting their CPU speed. At some point, there were just more important things to talk about.

If someone tells you “have you tried using Gemini instead”, you can respond with “different geniuses with the same bad instructions will still do the wrong work”.