Meta’s first specialized application is here with Llama 4 – but is more profitable than performance or business orientation

Facebook -Mother Company Meta Platforms, Inc. have Officially launches its own, free self -AI application, This move aimed at delivering a more personal and integrated AI experience of mobile devices, network and Ray-Ban Meta Smart Glasses.

The application is Available to iOS via Apple App Store and on the network – Without mentioning when a version of Android can come.

Powered by a version of his new, Dividing quasi code LLAMA 4 mixture of experts and considerations of the familyThe new Meta AI application focuses on learning the preferences of users, maintaining the conversation context and providing a hassle -free interaction with voice. It requires an account at Meta Products to enter, although users can log in to their existing Facebook or Instagram accounts.

Comes before the start of Llamacon 2025The first Meta AI Developer Conference, which is being held this week at its Campus office headquarters in Palo Alto, California, focused around her family Llama Model Family and General Ai Developer Tools and Advances.

With the rise of more AI Model competitors in the open code and own domains-all from Openai with Chatgpt to Google with its Gemini 2.5 Model Family and less known (at least to Western audiences) as brands such as brands Alibaba’s new Qwen 3 -Meta is willing to show the strength and capabilities of its own, internal models Llama 4.

He also strives to make the case of third -party software developers that Llama 4 is a powerful and flexible Open (ISH) Source Model family that they can trust to build their businesses at the top of. However, with this new launch of the Meta AI application, I am not sure that this is the most successful example. More about it below.

Text, image and voice outside the box-with the editing of documents

The Meta AI app is a new way for users to interact with AI AI AI beyond existing integrations with WhatsApp, Instagram, Facebook and Messenger.

This allows users to make natural, voice conversations with AI, edit and generate images and detect new cases of use through a cure Discover issue, including promotions and ideas shared by the Community.

Along with the traditional text interaction, Meta AI now maintains voice functionality, while being multitudinal. The early full duplex voice demonstration allows users to experience natural, flowing conversations in which AI generates speech directly, not just reading a text aloud.

However, the demonstration does not have access to real -time web information and may show accidental technical inconsistencies. Voice features, including the demonstration of full duplex, are currently available in the US, Canada, Australia and New Zealand.

On the Meta.Ai network, it is designed to reflect a mobile experience, offering voice interaction, access to the Discover issue and an improved tool for generating images with improved style, mood and lighting controls.

The web version seems particularly powerful and capable of creating images, with many predefined styles and ratios of countries to choose from. In my short practical tests with the mobile application, image creation tools seemed far more limited and I was unable to find a way to switch the sides. In both shapes, the image quality was far lower than the dedicated and rival AI image generators such as Midjourney or Native Image Generation of Midjourney or Openai.

Meta also tests a wealthy editor of documents and functions for analyzing documents in selected countries.

Discover what other users do and create with AI

A distinctive feature of the application is its Discover section, available by sliding from the main chatbot interface, where users can view and remix prompts, ideas and creative results shared by others.

This issue emphasizes how people use Meta AI to brainstorm, write, analyze content on social media, create stylized images and explore playful concepts-as a pixel-art design or search for satellites generated by AI.

Creator publications include both text prompts and image results, which gives others the starting point to experiment with AI in new ways. It also coincides with A recent monitoring of the technical journalist Alex Kantovitz (Great Technology) in LinkedIn Post The fact that AI is constantly replacing social media as a means of detecting entertainment and content for an increasing number of users.

This dynamics of peer sharing is aligned with Meta’s intention to make AI not only useful but also culturally engaging, offering a social layer of what is traditionally interacting one in one.

Seeing the future

For users of Meta Glasses Augmented Reality Ray-Ban, the Meta AI app replaces the previous meta visual application.

Existing pairs of devices, settings and media content will automatically migrate when updated.

This integration enables users to move from interaction with their glasses to the application, maintaining the history of conversations and access between devices and network, although the application cannot yet be initiated in the application and resumption of glasses.

Memory and personalization

Personalization is at the heart of the new Meta Ai Experience.

Users can instruct the Meta AI to remember certain interests and preferences, and the assistant also draws from user profiles and a history of meta platform engagement to adapt the answers.

This feature is currently available in the US and Canada. Users who connect their accounts to Facebook and Instagram through the Meta Account Center can benefit from a more in -depth personalization.

When I downloaded the app to try it, it automatically offered and pre -filled my input into the Instagram account.

Virtually

My initial tests on the Meta AI App interface reveal both the impressive functionality of Llama 4 and its ongoing restrictions on daily tasks.

On the one hand, the assistant is able to generate useful answers, offers analysis and tips and quickly generates images.

However, some interactions reveal severe restrictions that are most resolved in other AI applications and large language models (LLMS) that feed them behind the scenes.

In one case, Meta AI initially violated the number of “m’s in the word” mom “, adjusting only after being prompted to review its answer.

A similar model arose when the letter “R” in “Strawberry” arose, where first answered with 2 before adjusting to 3 after further clarification.

Another answer incorrectly evaluates which number is more large between 9.11 and 9.9, a task involving a basic decimal comparison.

These moments emphasize the restrictions of the model when it comes to details in short factual tasks for reasoning – a well -known area in which even advanced languages ​​models can fall apart.

The assistant’s ability to recognize mistakes, explain his reasoning, and offer transparent adjustments reflects the progress to more interactive and self-recording experiences.

But overall I can’t recommend being used for jobs right now.

Speaking of a wider strategy in video and audio interview with AI focused YouTuber and Podcaster Dwarkesh PatelMETA CEO Mark Zuckerberg stressed that personalization and trouble -free interaction with low latency are priorities for the development of AI of the company.

https://www.youtube.com/watch?v=ryxeqbtuvl0

“If a few years quickly forward, I think we’ll just talk to AI all day about different things we’re wondering,” Zuckerberg said.

He emphasized the focus of the company on construction systems, which are fast, local multimodal and integrate deeply into everyday life.

Zuckerberg also discussed Meta’s AI approach with open source, noting that Llama 4 models were designed to balance efficiency, intelligence and accessibility.

He has confirmed that the Meta Meta -Code’s Meta commitment aims to provide broad innovation while maintaining US leadership in the standards of the AI ​​model, especially in providing values ​​and the integrity of the system.

AI daily?

As a meta to compete on an increasingly crowded AI, launch of the Meta AI app marks a significant step towards turning intelligent, personalized assistants, part of everyday life for millions of users worldwide.

With the active feedback of consumers and the growing storage of shared use through the Discover issue, Meta apparently invests in an ecosystem, where AI develops in tandem with community commitment and real -world requirements.

While the Meta AI app is designed primarily for users, its launch has wider consequences for business in each sector.

Meta, with an audience of Nearly 4 billion users worldwide Within its applications and hardware products, there is a scale to transfer the public expectations for technology fundamentally.

Even if only a small part of its users withdraw and engage in the Meta AI app, it will introduce millions-probably hundreds of millions of non-technical users on regular, casual interaction with AI and will show the options for conversational interaction in text and voice, quickly generate images and problem solving.

This basic exposure is likely to accelerate a change in what people expect not only from user applications but also from jobs, service providers, retailers and any type of merchant or provider with whom they interact.

When people become accustomed to a personalized, conversational AI, who can understand the context, anticipate the needs and support creative or information tasks, they will expect and require similar functionality everywhere – from their own jobs and those businesses and businesses from which they buy.

Companies that do not offer affordable, responsive experiences, managed by AI, risk feeling outdated or non -responding compared to what users are increasingly accepting for granted.

In fact, the new META application may not simply compete with other AI offers; It can redefine the main line for digital interaction standards in industries.

Businesses, regardless of the size or sector, will need to rethink how they include AI in customer experiences, service channels and even internal operations if they want to meet the new cultural expectations that it is a wide knowledge of consumers with AI.