Does OpenAI have a sustainable business model? ChatGPT usage is down (from its initial astronomical numbers), but – more importantly – by the numbers, the power users are engineers. When it comes to coding, GPT-4 makes good engineers great and it makes bad engineers better.
Last week, Meta launched Code Llama: an open source, AI model specifically fine-tuned for coding tasks. This is interesting as the bulk of GPT-4’s commercial power users now have a free alternative. Code Llama comes in various sizes (7B, 13B, and 34B parameters) and can be configured to run in a wide range of computational environments. It can generate code in multiple languages – including Python, Java, JavaScript, C#, and Bash – and it’s available for both research and commercial use.
Beyond code generation, Code Llama can explain code in natural language, serving as a tool for software developers. Additionally, it can be embedded in business applications to auto-generate and execute code snippets based on natural language prompts.
In terms of performance, Meta’s research positions Code Llama ahead of GPT-3.5 on multiple code benchmarks. Independent tests by Snowflake and AI startup Phind further validate its capabilities, with some results nearing GPT-4 performance in specific tasks.
Is Code Llama a threat to OpenAI? Just remember: “free” is very pro consumer!
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it.