The new tool is a fine-tuned version of LLama2 trained specifically to generate and discuss computer code. Code Llama could be particularly useful for small fintechs traditionally underserved by big-tech AI providers
Meta AI announced the launch of ‘Code Llama,’ a community-licensed artificial intelligence (AI) coding tool built on the Llama2 large language model (LLM) that can use text prompts to generate and discuss code, on Aug. 24.
Code Llama is released for both research and commercial use under the same community license as Llama 2. That means it can be used by businesses and individuals for free with proper attribution.
The tool is expected to boost developers’ productivity, speed up workflows, and make coding more accessible. In particular, it can become a viable alternative to the costly GitHub’s Co-Pilot (built with ChatGPT technology).
According to the dedicated blog post, Code Llama has several variants. One model is fine-tuned for general coding in a number of languages (including Python, C++, Java, PHP, Typescript, C#, Bash and more), while another is fully dedicated to Python. Two of the tool variants have in-built fill-in-the-middle (FIM) capability, allowing it to support tasks like code completion. In addition, Code Llama can be leveraged for debugging purposes.
While two Code Llama model variants are focused on understanding, explaining, and discussing code, Code Llama Instruct is recommended for actually generating code. The models are presented in different parameter sizes to be able to operate in different technical environments.
The open-source AI coding tool introduction by Meta could be a game-changer for businesses and individuals who require LLM models for coding purposes but can’t afford the costly options already present on the market. Some of the traditionally underserved segments are small-scale fintech startups, Web3 developers, digital asset service providers, etc.