On August 24, 2023, Meta released Code Llama—an AI model specialized for writing code, completely free and open-source.

GitHub Copilot charged $10-19/month. Code Llama was free, runnable locally, and customizable. For developers, this was a game-changer.

What Code Llama Offered

Code Llama came in three sizes, each with different capabilities:

7B: Fast, runs on consumer hardware, good for code completion 13B: Balanced performance, suitable for most coding tasks 34B: Most capable, approaches commercial AI coding assistants

All three were released with full weights for free commercial use.

Specialized Variants

Meta released specialized versions:

Code Llama - Python: Fine-tuned specifically for Python programming Code Llama - Instruct: Trained to follow natural language instructions for coding tasks

This specialization made the models more effective for specific use cases than general-purpose models.

How It Compared

Against commercial alternatives:

vs. GitHub Copilot: Comparable code completion, free, runs locally vs. GPT-4 for coding: Less capable on complex tasks, but faster and free vs. Claude for coding: Similar trade-offs—less context, but open and customizable

For many developers, "good enough and free" beat "slightly better but paid."

The Use Cases

Developers adopted Code Llama for various purposes:

Code Completion

Real-time suggestions while coding, similar to Copilot but running locally with full privacy.

Code Generation

Natural language descriptions turning into working code—"create a function that validates email addresses."

Bug Finding

Analyzing code to identify potential issues, security vulnerabilities, or performance problems.

Documentation

Generating comments, docstrings, and README files from code.

Learning Tool

Students and junior developers using it to understand coding patterns and best practices.

The Local Advantage

Unlike API-based tools, Code Llama could run entirely on your machine:

Privacy: Your code never leaves your computer No internet required: Code offline without issues No usage limits: Use as much as you want Customization: Fine-tune on your codebase's style

For enterprises with strict security requirements, this was crucial.

The Developer Ecosystem

Within weeks, developers built tools leveraging Code Llama:

IDE extensions: Plugins for VS Code, JetBrains, Vim Code review tools: Automated PR analysis Documentation generators: Auto-creating docs from code Learning platforms: Interactive coding tutors

The open-source nature enabled rapid innovation.

The Impact on GitHub Copilot

Code Llama put pressure on commercial alternatives:

Pricing questions: Why pay when free alternatives exist? Feature differentiation: Paid tools needed clear advantages Integration matters: GitHub's ecosystem integration remained valuable

GitHub Copilot didn't die, but it faced real competition for the first time.

Where Are They Now?

Code Llama continues being widely used, though newer models like Llama 3 with coding improvements and specialized tools like Cursor have emerged.

The impact was less about Code Llama specifically and more about proving open-source AI coding tools could compete with commercial offerings.

Today, developers have multiple free options for AI coding assistance. Code Llama helped make that reality.

August 24, 2023 was the day AI coding went from "premium feature" to "freely available tool" for all developers.