Meta Releases Code Llama AI Developer Assistant

Facebook parent company Meta announced on August 24 that it is releasing Code Llama, a large language model (LLM) for use by programmers and developers.

The company said that Code Llama can generate and discuss code, increase workflow speed, and lower barriers to entry for new programmers.

Meta added that the new product is a customized version of Llama 2, its large open source language model. The new product is available for research and commercial use under the same license as Llama 2, which opened in July.

Meta claims that Code Llama can generate code as well as natural language about code, and that the tool can also take both types of information as input. The company also stated that, in addition to generating code and content, Code Llama can be used for debugging. Meta said it supports popular programming languages including Python, C++, Java, PHP, Typescript (Javascript), C#, Bash, and more.

The company said it is releasing three versions of Code Lama, each trained with different parameters and requirements. According to the latest Meta announcement, the smallest version of 7B can be served on a single GPU. On the other hand, the largest 34B model produces the best results and offers the best coding assistance.

The company also announced a Python-centric version of Code Lama, as well as an instruction-centric version of Code Lama, which is excellent at understanding human prompts.

In the blockchain industry, some have suggested that AI can be used to write smart contracts that underpin popular decentralized applications (dApps) and DeFi platforms, and to discover vulnerabilities in existing smart contracts.

While Meta has not announced support for popular smart contract languages such as Solidity, other projects such as ChainGPT and Decrypted are working on providing similar features that allow developers to generate such code with an AI tool.