Codeium: The best GitHub Copilot Free Alternative

This week I discovered Codeium: a very good alternative to GitHub Copilot. It’s free for individual developers. It’s very responsive and enterprises can also host it on their own infrastructure.

Let me tell you more.

You can also watch or listen the content of this article on YouTube.

GitHub Copilot vs JetBrains Full Line Code Completion

A couple of months ago I made a video explaining what I expect from an AI Coding Assistant and why I thought GitHub Copilot was superior to the competition. In a few words, GitHub Copilot was the best tool at performing inline code completion. Not only because it was extremely responsive but also because its suggestions took into account the entire context of my project.

That means it’s sometimes capable to suggest the code for an entire class just by looking at its name and what you created previously in the project.

Now before I talk about the free alternative I discovered, let me tell you how I found it.

On April the 4^th, JetBrains covered its new Full Line Code Completion feature in a blog post. I was immediately hooked, but I tell you that it was a disappointment.

Once I enabled it, coding became a pain. I am used to getting meaningful multi-line code suggestions and working out of single-line code suggestions was unbearable. Then, I realized that it’s meant to work in conjunction with the JetBrains AI Assistant but since I prefer GitHub Copilot and it’s not compatible with it, I just gave it a pass.

That said the idea of having a coding assistant that can run locally and does not require an additional license was enticing. And that’s how I found Codeium!

Codeium vs GitHub Copilot

Codeium is the equivalent of GitHub Copilot. However, it’s free forever for individuals with unlimited usage. Enterprises can opt to run it on their own infrastructure. And it also has some experimental features which I find very interesting.

Now the question is whether the responsiveness and quality of code suggestions is inferior or superior to Copilot.

In terms of responsiveness, I have no doubts. Codeium is on par with GitHub Copilot. I was really surprised because I never expected a free tier to have that quality of service. I have literally nothing to add on this point. Codeium is lightning fast.

Now, what about the quality of code suggestions?

I am working on writing my own version of Git as part of a CodeCrafters challenge and I could see no major difference. If you have never heard about CodeCrafters, it’s a learning platform where you can practice writing complex software. If you want to try it out, make sure to use my affiliate link to get a huge discount: https://app.codecrafters.io/join?via=marcolenzo

The chat mode is fine as well. The responses are drier compared to GitHub Copilot. However, you can leverage the GPT4 model if you have a paid seat. That said, you should know from my previous video that I don’t like to overuse prompts or chat, so I personally don’t care much about this aspect.

The additional feature I like is context pinning which allows you to mark a repository, directory, file or function as something Codeium should keep in higher consideration when suggesting code. This helps Codeium generate code that is more aligned with your intents. This is still an experimental feature in addition to the existent opened files context and repo wide context.

Is Codeium good for You?

In conclusion... which one is better?

You know the answer! It depends.

Both tools are not perfect.

I personally decided to stop my GitHub Copilot subscription so I can test Codeium in more depth for at least one month. My gut feeling is that our perception of the quality of suggested code will change according to the programming language and type of problem we are trying to solve.

However, I have no doubt that if you are someone who codes on the side for fun, Codeium is the right option. I cannot imagine any valid reason why you would spend $10 a month when you can get something similar for free. Students and teachers need to decide for themselves since GitHub Copilot is free for them as well.

If you work in a professional environment, the answer is way more difficult. I find appealing the ability to deploy Codeium in an air-gapped environment. I know some companies are not comfortable with their code being shipped to the cloud, despite GitHub making a huge effort reassuring us in its Trust Center.

I also know there are other solutions like LMStudio that allows you to download and run a model right on your machine, but that requires very beefy hardware and I don’t think it’s something we can all afford.

That said, I hope I ‘ve given you something to try tonight or this weekend in your free time! Don’t forget to like and share so that everyone can learn something new!

Codeium YouTube Video