Hi Reader, āLast week, I encouraged you to experiment with different LLMs, since thereās no one model that is superior across all use cases. Specifically, I suggested you try using Chatbot Arena, which allows you to chat with multiple models at once. Itās completely free, but has two significant disadvantages:
Today, I want to offer you a better method for experimenting with LLMs, and itās the method I personally use. But first, we need to talk about LLM pricingā¦ š· Whatās the cheapest way to access LLMs?If youāre familiar with ChatGPT, you probably know that it offers a limited Free plan and a $20/month Plus plan. But did you know that thereās a more flexible, affordable way to access the paid features of ChatGPT? If you access LLMs via API, you actually pay per ātokenā, which is roughly equivalent to a word. The LLM I use the most, GPT-4o mini, costs only $0.15 per million āinput tokensā (what you send to it) and $0.60 per million āoutput tokensā (what it sends back). That means most of my conversations cost fractions of a cent! And when I want to use a more powerful model (like Claude 3.5 Sonnet), I pay perhaps $0.10 for a conversation. In summary, accessing LLMs via API means that you can use whichever models you need, and only pay for the tokens you use. Win-win! š How to access LLMs via APIIf you like the idea of accessing LLMs in a flexible, affordable way, thereās only two things you need to get started:
An āAPI keyā is just a unique identifier that allows you to access a web service. For example, my API key from OpenAI allows me to use their models, and they bill me for the usage. Once you have an API key, you need an interface for chatting with the LLM. My preferred interface is TypingMind: All you have to do is give TypingMind your API key, and it will communicate with the LLM on your behalf! I love TypingMind because itās so easy to switch between different LLMs while still tracking your conversation history in one place. Itās also got a ton of power user features: AI agents, plugins, voice mode, document upload, cloud sync, etc. ā© Action itemsThis week, I want you to try using an LLM via API. Here are the steps:
Next time, I'll give you tips for how to get the best results from LLMs!
Tuesday Tips is on pause during the holiday break, so I'll see you in 2025! - Kevin P.S. Your workouts are a work of art (27-second video, turn on audio!) Also: behind the scenes. ā Did someone AWESOME forward you this email? Sign up here to receive weekly Artificial Intelligence tips! |
Join 25,000+ intelligent readers and receive AI tips every Tuesday!
Hi Reader, The Python 14-Day Challenge starts tomorrow! Hope to see you there š¤ š Tuesday Tip: My top 5 sources for keeping up with AI I'll state the obvious: AI is moving incredibly FAST šØ Here are the best sources I follow to keep up with the most important developments in Artificial Intelligence: The Neuron (daily newsletter) My top recommendation for a general audience. Itās fun, informative, and well-written. It includes links to the latest AI news and tools, but the real goldmine is...
Hi Reader, Before todayās tip, I wanted to give you a heads up: Tomorrow, Iāll be launching something brand new! Watch out for the announcement š š Tip #53: How to get great results from AI models through prompting In the year after ChatGPT was released, I remember noticing two new trends: Articles about āprompt engineersā being hired for hundreds of thousands of dollars just to write prompts Endless guides promising to teach you the secrets of writing the perfect ChatGPT prompt My takeaway...
Hi Reader, Over the past 50 tips, Iāve touched on many different topics: Python, Jupyter, pandas, ML, data visualization, and so on. Going forward, Iām planning to focus mostly on Artificial Intelligence. Iām announcing this so you know what to expect, and I know what to deliver! š Iāll also try to make the tips shorter, so that they're easier to digest on-the-go. Finally, I plan to include an āaction itemā each week, so that you can practice what youāre learning. I hope you like these...