Code Llama Models by AI at Meta

About

Code Llama is a family of large language models developed by Meta AI, designed for code generation and understanding tasks. Building upon the Llama 2 architecture, Code Llama offers variations to accommodate diverse needs, featuring model sizes with 7B, 13B, 34B, and 70B parameters that adjust complexity and capability. The family comprises the foundational model for general code tasks, Code Llama - Python for tasks specific to the Python language, and Code Llama - Instruct, which is fine-tuned for interpreting natural language instructions. Trained on extensive code and related data, these models excel in code completion, debugging, and generating code from natural language prompts. Larger models tend to provide enhanced performance, albeit at the cost of increased computational demands 12.

Specifications(16 models)

Available From(9 providers)

Baseten API

DeepInfra

Fireworks AI

IBM watsonx

Microsoft Foundry

NVIDIA NIM

OctoML (Deprecated)

Replicate API+1 more

Pricing

Frequently Asked Questions

What is Code Llama used for?

Code Llama is used for coding and structured outputs. The family description and listed model capabilities point to those workloads as the best fit.

How does Code Llama compare to Claude Fable?

Code Llama by AI at Meta is strongest where you need coding, while Claude Fable by Anthropic is the closest related family to check for vision and multimodal work. Code Llama has 16 listed variants and reaches up to 100k context, while Claude Fable reaches up to 1m context, so compare the specs and pricing tables before choosing a production model.

Which Code Llama model should I use?

For the lowest listed input price, start with CodeLlama 7B through Replicate API at $0.05/1M input tokens. For the most capable/latest local choice, evaluate CodeLlama 34B with 100k context and structured outputs.

Related Model Families

Claude FableClaude 4.8

Claude 4.7

Claude 4.6

Claude 4.5

Claude 4

Models(16)