Before discussing what LLM to use, I will start with what is an LLM? A Large Language Model (LLM) is an artificial intelligence system trained on vast amounts of text data to understand and generate human-like language. You can chat with an LLM to learn about…pretty much anything. It can generate written content. It can proofread, summarize data, and more.
There are dozens of major LLMs and hundreds of variants that are based on open-source models. There are so many options that I could write a book. Unfortunately, by the time I completed the book, the information would be out of date. Instead I will focus on five mainstream options.
I have been working with the free editions of ChatGPT (OpenAI), Claude (Anthropic), Perplexity, Gemini (Google) and Copilot (Microsoft) for a while. Because I have been using it for so many things (professional and personal), I frequently run out of tokens. It is time to subscribe. But like streaming services, I don’t want to pay for them all.
Comparing LLMs
To evaluate, I started by asking all five models to compare themselves with the others. I also asked, “Which one would you choose?” Rather than having AI to write this post, (though I will likely ask all five models to proofread), I am summarizing the keys points that will help me make my decision:
- Perplexity currently provides the most current data. It will also cite it’s sources.
- Claude was universally identified as having the strongest reasoning skills.
- Chat GPT is considered the most versatile. It can assist with text, image, audio and video, while also having the most extensions and plug-ins.
- Gemini and Copilot’s are at their best when integrated with their respective ecosystems – Google and Microsoft.
- All LLMs offer a free version and an entry level paid version for $20/month. Claude has an additional tier for $100/month, ChatGPT has a professional edition for $200/month, and Copilot has multiple tiers going to about $five0/user/month.
- Gemini is the only model that did not choose itself. It leaned toward Perplexity. While Perplexity considers itself unmatched for fact-based work, it acknowledges that for uses, other LLMs may be better suited.
Selecting an LLM
Just as I was narrowing down the options, I stumbled upon OpenRouter. Instead of a monthly subscription, it allows you to purchase tokens to access to a long list of LLMs. With the data OpenRouter collects, it can optimize the selection of an LLM. This is a great option for developing integrations and automations, but not ideal for an individual doing research or generating content.
Back to my five options. I can easily eliminate Gemini and Copilot. For my purposes, I do not require seamless integration with Google or Microsoft. I can continue to use all the models for free, so if I want to do something that is best suited for a particular LLM, I will always have that option.
So which LLM will receive my dollars? For now, ChatGPT (Plus Plan). Versatility is the deciding factor. I think it will be “good enough” for most of my needs but I also know I will need to reevaluate often. Things are evolving quickly, and in a few months, I may reach a different conclusion.
If you want to learn more about LLMs, check out Hugging Face or LinkedIn.