Meta's Code Llama 70B is here to take the fight to GitHub Copilot

We have come a long way since we last saw Meta aiming to build an open-source ChatGPT contender, with a recent announcement on their official X page (formerly Twitter) they have introduced Code Llama 70B.

They dub it as “a new, more performant version of their LLM for code generation”; Let's see what they claim about it.

Suggested Read 📖

Code Llama 70B: What to Expect?

a screenshot of code llama in action — Official image that shows how Code Llama works.

Released under the same license as the Llama 2, Meta asserts that this license makes it possible to provide Code Llama 70B for both research and commercial uses.

With this release, Code Llama now features a much larger 70B parameter model that, in theory, should provide users with a better code output performance.

For those who are not familiar with Code Llama, it is a large language model (LLM) that can process text prompts to generate code, Meta says that it has the potential to make workflows faster and more efficient.

It can also be used to perform debugging, code completion, and has support for many popular programming languages such as Python, C++, Java, PHP, and more.

Think of it as ChatGPT, but for code! 😃So, in other words, a GitHub Copilot contender.

On the launch of Code Llama 70B, Mark Zuckerberg, CEO of Meta, had this to add:

We're open sourcing a new and improved Code Llama, including a larger 70B parameter model. Writing and editing code has emerged as one of the most important uses of AI models today.

The ability to code has also proven to be important for AI models to process information in other domains more rigorously and logically. I'm proud of the progress here, and looking forward to including these advances in Llama 3 and future models as well.

Furthermore, they have made available three variants of Code Llama 70B that cater to different needs:

CodeLlama-70B, which is the foundational code model.
CodeLlama-70B-Python, which is tailored for Python and provides additional support.
CodeLlama-70B-Instruct, which has been fine-tuned for understanding natural language instructions.

For the CodeLlama-70B-Instruct variant, Meta claims that it scored a 67.8 on HumanEval, and that it makes it “one of the highest performing open models available today”.

At the time of writing, there was no new official blog on this, however, they did update their older blog on Code Llama. You can also check out the official announcement over at X.

It is great to see that there is some really robust competition to GitHub's Copilot. Code Llama 70B is in a great position to cater to the many developers, educators, and students out there who would like to take advantage of such an AI-powered tool.