The CEO of Hugging Face, the platform that hosts most of the world’s open-source AI models, made a striking claim recently. Clem Delangue said that most enterprises do not want the most powerful frontier AI models. They want open models they can control, customise, and run themselves at a reasonable cost.
This goes against the narrative you hear in tech news, where the focus is always on which company has the smartest model. But Delangue’s observation matches what many Irish small business owners are discovering: the most powerful AI is not always the most useful AI for everyday business tasks.
The Difference Between Frontier and Open Models
Frontier models are the headline-grabbers. ChatGPT, Claude, Gemini. These are built by the biggest tech companies, cost enormous amounts to develop, and are only available through their creators’ servers. They are extremely capable but come with ongoing per-use costs and limited control over how they work.
Open models, by contrast, have their internal weights and code published openly. Anyone can download them, run them on their own hardware, modify them, and use them without paying per-query fees. The most capable open models are not as powerful as the very best frontier models, but they are catching up fast. Models like Meta’s Llama, Mistral’s offerings, and now Thinking Machines’ Inkling are closing the gap.
What This Means for Your Business
For the average Irish small business, the choice between open and frontier AI comes down to a few practical questions:
What are you actually using AI for? If you need help drafting emails, summarising documents, answering customer questions, or generating marketing copy, you do not need a frontier model. An open model can handle these tasks perfectly well, and you can run it for free on your own computer.
How much do you use it? Cloud AI charges per token. A business that uses AI heavily for customer support, content generation, or data processing can run up significant monthly bills. Open models, once you have the hardware, cost nothing extra to run.
Do you care about data privacy? Every query you send to a cloud AI service leaves your business. If you are processing customer data, financial information, or anything confidential, running an open model on your own equipment keeps everything under your control.
When Frontier Models Still Win
There are situations where frontier models make sense. If you need cutting-edge software engineering, complex mathematical reasoning, or highly specialised analysis, the very best models still outperform open alternatives. But these use cases are rarer than the AI industry would have you believe.
For most everyday business tasks, the gap between open and frontier models is small enough that it does not matter, and the practical advantages of open models, cost, privacy, and control, often outweigh the slight difference in output quality.
A Practical Approach for Irish Businesses
Here is a straightforward way to think about this. Start with free tools. Try out a service like ChatGPT or Claude for a month to see what AI can do for your business. Learn the basics of prompting and understand what tasks AI handles well.
Then, if you find yourself using AI regularly and spending noticeable money on it, look into open alternatives. Try running a model through Ollama, a free tool that lets you download and run models on your laptop. See if the results are good enough for what you need day to day.
Many businesses will find that open models handle 80 percent of what they need at a fraction of the cost. The frontier is exciting, but for most of us, the real value of AI is closer to home.