Meta Unveils New AI App to Compete with ChatGPT
On 29 April 2025 at Meta’s inaugural AI developer conference, LlamaCon, Meta has launched a suite of AI products.
These products aimed at challenging OpenAI’s ChatGPT, including a new standalone AI assistant app, a developer-friendly Llama API, and a set of safety tools for open-source AI models.
Key Announcements
Meta AI App
Meta new AI App is powered by Meta’s latest Llama 4 model and offers personalized assistance by remembering user preferences and context.
This app has standout feature is the “Discover feed,” where users can share and explore how others use AI.
The app also integrates with Facebook and Instagram accounts to enhance personalization.
The app supports voice interactions through an experimental full-duplex mode for more natural conversations.
Llama API: Accelerated AI Development
Meta has previewed its Llama API, designed to simplify the integration of Llama models into applications.
Llama API offers one-click key creation, interactive playgrounds, and tools for fine-tuning models.
Llama API is now available in limited preview. This new tool enables developers to build applications powered by Meta’s Llama family of large language models, starting with Llama 3.3 8B.
Notably, Meta assures developers that user data won’t be used for model training, and self-hosting remains an option for data privacy.
Through partnerships with Cerebras Systems and Groq, Meta promises lightning-fast inference — up to 18 times faster than standard GPU solutions, reaching speeds of 2,600 tokens per second for some Llama 4 models.
Enhancing AI Safety: Llama Guard 4 and More
To address safety and security in AI applications, Meta introduced new tools for the open-source community: Llama Guard 4, LlamaFirewall, and Prompt Guard.
These tools are designed to help developers evaluate and mitigate potential risks associated with AI models.
Meta also introduced CyberSOC Eval and AutoPatchBench, two new tools for assessing AI system defenses, both available as part of CyberSec Eval 4, its updated open-source cybersecurity benchmark suite.
CyberSOC Eval measures the efficacy of AI systems in security operation centers, while Autopatch Bench evaluates AI systems’ ability to automatically patch vulnerabilities in native code.
Background
Meta has steadily expanded its AI efforts since launching Meta AI in September 2023.
As of January 2025, Meta AI reached 700 million monthly active users, up by 100 million in just a month.
With these new tools and collaborations, Meta is strengthening its position in the AI race — offering powerful consumer applications and robust developer infrastructure to shape the next wave of AI innovation.
News Gist
At LlamaCon, Meta launched a new AI assistant app with a social “Discover” feed, to rival ChatGPT. Meta also unveiled the Llama API for faster AI integration, and introduced safety tools like Llama Guard 4 .
Zhipu Launches RoboOS 2.0 and RoboBrain 2.0 to Power Smarter Robots
ETH Zurich’s AI-Powered Robot Showcases Exceptional Agility in Badminton Play
World’s First Humanoid Fighting Competition Held in China
Amazon Unveils ‘Vulcan’, Robot for Warehouses
Hyundai and Persona AI To Develop Humanoid Welding Robots
