Here's how to use Claude for free

Here's how to use Claude for free with some hacks to get the absolute most out of a free account on what you can actually do with it for free. So to be clear, the free version runs on Sonnet 4.6. That's their latest mid-tier model. So, no, you don't get Opus, the most intelligent one. But honestly, for almost everything like your everyday tasks, you don't need it and Sonnet is more than enough. In terms of features, you can upload up to 20 files in one chat, up to 500 megabytes each. So you can do some serious number crunching, data analysis, document analysis, PDF, spreadsheets, images, word documents, etcetera. You can throw all kinds of things at it. If you give all your bank statements and ask it to create a full budget report spreadsheet, that's free. And yes, that file creation used to be a paid feature. You now get that for free and Claude is brilliant at creating PDFs, Word, Powerpoint, Excel documents, etcetera. Web search is on by default now. So, it's not stuck in the past. You get artifacts where it builds these little documents or tools in the right hand side pane, they're very useful. There's also memory that went free earlier this year. So, even on a free plan now, it remembers things across your chats and it gets to know you over time. But as you use these features, you will absolutely encounter usage limits all the time. Anthropic doesn't publish the exact numbers, but it's about 15 to 40 messages every five hours. And you're gonna run out of that really quickly. But it's this rolling window, okay? So it doesn't reset at midnight, it's a rolling window that starts from your first message. Anyway, there are four ways that will make a huge difference in how much you get out of that free usage. The first one is start a fresh chat for every new task. That means Claude doesn't have to read over lots of previous conversation just to respond. Secondly, use projects. This is a big one. You get 5 projects in a free account and a project is a folder with memory and anything you upload to it gets cached, meaning that it only has to read it once and then, it remembers it. So, you can keep asking questions about the same documents and it doesn't burn through loads of tokens re-reading those documents, okay? The third thing is just to be efficient with your input and output, so be specific and efficient in your prompts, especially with attached files because like I've said, reading through files burns a lot of tokens. So, if only needs pages 2 and 3 of a 20 page document, extract pages 2 and 3 and just give it them. But also with the output, if you only need 200 words, tell Claude that, so it doesn't waste tokens giving you 500 words. And then the fourth thing is match the model to the job. So for simple stuff, for most stuff, drop down to Haiku that will use a lot less usage and also lower the effort level in the model picker that will bring down your usage a lot as well and save the clever slower settings when you really need them for those more complex or multi step tasks. Also turn off web search and connectors when you don't need them because they can be really token hungry. Also a bonus tip from Anthropic, don't bother being polite with please and thank you, it just wastes tokens.