GPTZero detects Meta’s new Llama AI with 99% accuracy

For the past year,  students and professionals alike have been using AI tools like ChatGPT to improve their writing and tackle tasks from mathematics to language translation to scheduling. While ChatGPT is the most well-known AI model, other large language models (LLMs) have  joined the race, like Anthropic's Claude, Google's Gemini, and Meta's open-source AI model Llama. 

What is Llama?

Llama is Meta’s AI language model that has been making waves for its sophisticated capabilities and impressive performance. It’s been touted as one of the most advanced AI models to date for its ability to generate text that sounds human. Llama 3.1, the latest model released on July 23, 2024, supports even more languages, allows for longer text prompts, and integrates with more software services.

Speaking about Llama on a recent quarterly earnings call, Meta CEO Mark Zuckerberg said:

“I think we are going to look back at Llama 3.1 as an inflection point in the industry where open-source AI started to become the industry standard, just like Linux is. My view is that Open Source will be safer, will enable innovation that improves all of our lives faster, and we’ll also create more shared prosperity.”

But despite Zuckerberg's vision for Llama, educators, journalists, and technologists have expressed concern if Llama’s ability to sound human will make it harder to know when a source of content is AI generated. Potential perilous cases of AI generated content in the past have led to AI search engines collapsing under “fake” AI articles. AI detection tools like GPTZero are doing our part to preserve what’s human by keeping pace and refining our tool to the latest models  as mainstream philosophies around AI continue to evolve. 

Can GPTZero Detect Llama? 

GPTZero can detect Llama as accurately as it can classify other GPT-based models, from ChatGPT to Claude to Gemini. GPTZero has previously performed superior to other AI detection competitors based on independent studies. TechCrunch calledGPTZero the “best and most reliable” AI detector for catching AI writing 5 out 7 times in their 2023 experiment, the highest of the tools they evaluated. 

GPTZero vs. Llama 3.1: How does AI detection match up

Recently, GPTZero achieved a 99% accuracy rate on texts that were produced through Llama’s 3.1 AI. 

We owe our success to three factors:

1. Advanced algorithmic precision: We constantly update our models to analyze subtle nuances in language, syntax, and context.
2. Continuous learning: We stay abreast of the latest developments in AI – including with Llama –  and continuously update our models to keep the pace.
3. Robust training data: Every new AI tool that crops up allows us to extend and diversity our training data sets.

Try GPTZero’s AI detector for free. Sign up for deeper scans and writing tools.

What’s next for AI detection

AI-generated content is getting more sophisticated. We aren’t the only team to ask ourselves if one day, it will outpace our ability to detect it. But we’re committed to staying at the forefront of AI innovation, and we’re proud that GPTZero identifies even the most advanced AI-generated content with the highest degree of accuracy. 

Better AI allows us to get better at what we do. So when it comes to testing our models against more sophisticated AI outputs, we say, “Challenge accepted.”