Meta’s free Code Llama AI programming tool closes the gap with GPT-4

Meta’s latest update to its code generation AI model, Code Llama 70B, is “the largest and best-performing model” yet. Code Llama tools launched in August and are free for both research and commercial use. According to a post on Meta’s AI blog, Code Llama 70B can handle more queries than previous versions, which means developers can feed it more prompts while programming, and it can be more accurate.

Code Llama 70B scored 53 percent in accuracy on the HumanEval benchmark, performing better than GPT-3.5’s 48.1 percent and closer to the 67 percent mark an OpenAI paper (PDF) reported for GPT-4.

Built on Llama 2, Code Llama helps developers create strings of code from prompts and debug human-written work. Meta simultaneously launched two other Code Llama tools last fall, Code Llama - Python and Code Llama - Instruct, which focused on specific coding languages.

Code Llama 70B is available on three versions of the code generator and is still free for research and commercial uses. The large model was trained on 1TB of code and code-related data. It is hosted on the code repository Hugging Face, which offers access to GPUs to run AI models.

Meta said its larger models, 34B and 70B, “return the best results and allow for better coding assistance.”

Other AI developers released code generators in the past year. Amazon’s CodeWhisperer came out in April, while Microsoft harnessed OpenAI’s model to bring GitHub Copilot.