WhatsApp began testing AI-generated replies for business inboxes this week as part of Meta’s effort to automate customer messaging, according to the company’s Help Center. The feature allows businesses to generate suggested or automated responses within the WhatsApp Business environment, with controls to manage when and where AI replies are used.

The AI-generated reply feature for WhatsApp business inboxes is designed to assist companies in managing high volumes of customer messages by offering suggested or automated responses within the WhatsApp Business environment, according to WhatsApp’s Help Center. The feature is framed as an assistant tool rather than a fully autonomous chatbot, emphasizing controlled automation to improve response times for frequently asked questions.

Businesses can control when and where these AI replies are deployed, allowing for flexibility in automating parts of the customer communication process.

WhatsApp’s Help Center provides a step-by-step guide for setting up Business AI, directing users to access the feature through the Tools tab, then “Your Business AI,” followed by “AI replies.” Meta’s Facebook Business Help documentation also includes instructions for configuring Business AI on both Messenger and WhatsApp platforms. Meta describes the AI as capable of reading, understanding, and responding to customer messages, highlighting the company’s intent to integrate AI-driven automation across its messaging services. The official framing stresses governance and user controls to prevent unmoderated or unrestricted AI interactions.

The controls available to businesses extend to configuring AI replies as drafts, which require human review before sending, according to Hostaway’s Inbox AI Replies documentation, a third-party platform that supports WhatsApp automation. This approach suggests Meta is pursuing a model of configurable automation rather than a one-size-fits-all solution. Businesses can decide the degree to which AI handles the messaging workflow, which is particularly important for sensitive conversations that require human oversight. WhatsApp’s Help Center confirms that these inbox-level controls are designed to help companies tailor AI use to their operational needs.

The feature’s primary use case is to reduce response times in busy inboxes by quickly answering common questions. According to an analysis by Respond.io, businesses employing WhatsApp AI agents use them to greet customers, ask qualifying questions, interpret intent, and route conversations to the appropriate human agents or departments. Respond.io also notes that businesses need WhatsApp API access and a supporting platform to effectively manage conversations at scale. This aligns with Meta’s broader strategy to enhance WhatsApp’s utility for sales and customer support operations.

Additional automation features in the WhatsApp ecosystem include settings for AI agent selection, header and footer messages, stop keywords, and retry attempts, as demonstrated in a YouTube video showcasing WhatsApp AI agent configurations. Stop keywords allow users to disable AI responses and request human handling, while retry controls enable businesses to specify how many times the AI should attempt to respond if a message fails to send or process. These features indicate that automation is being implemented with multiple guardrails to maintain quality and user control.

Meta’s expansion of AI-enabled messaging features is evident across its platforms. Facebook Business Help documentation includes setup guides for Business AI on both Messenger and WhatsApp, and Instagram is also testing “smart replies,” according to publicly available information. This broader rollout reflects Meta’s effort to keep business communications within its suite of apps rather than relying on external tools. WhatsApp’s own Help Center materials confirm ongoing development and support for AI reply workflows.

Verified business use cases for WhatsApp AI agents include lead qualification by interpreting natural language and progressively collecting structured information such as product interest, budget, timeline, and location, according to Respond.io. Visito, a company specializing in WhatsApp AI agents, states that modern AI can understand conversational context, remember previous interactions, and solve problems more autonomously than traditional scripted bots. The business value centers on faster customer support, reduced manual workload, and improved lead handling efficiency.

The introduction of AI-generated replies in WhatsApp business inboxes is part of Meta’s larger automation initiative aimed at streamlining customer communication and enhancing the functionality of its messaging platforms for business users. Official documentation and third-party analyses confirm that the tools are designed with configurable controls to balance automation and human oversight.

Meta announced on Tuesday it has reserved up to 1 gigawatt of capacity from space solar startup Overview Energy, with the first orbital demonstration planned for 2028 and commercial power delivery expected by 2030. According to company officials, the solar-powered data centers will use satellites in geosynchronous orbit to provide continuous energy, aiming to cut energy costs for AI infrastructure by 30 percent.

The first orbital demonstration of Meta’s space-based solar power system is scheduled for 2028, with commercial power delivery expected around 2030, company officials said. The initial pilot project will include 25 megawatts of capacity and 2.5 gigawatt-hours of energy storage, according to details provided by Overview Energy, the space solar startup partnering with Meta. Meta has reserved up to 1 gigawatt of capacity from Overview Energy, with an agreement that grants the company preferred access to future capacity once technology milestones are met.

Meta’s solar-powered data centers will rely on satellites positioned in geosynchronous orbit roughly 22,000 miles above Earth, where the sun is continuously visible.

This arrangement allows the satellites to collect solar energy uninterrupted by weather, time of day, or seasonal variations, then beam the energy to collectors on the ground. The collected sunlight will be converted into electricity for distribution through Earth-based power grids, according to Overview Energy representatives.

The company emphasized that the data centers themselves will remain on Earth for easier maintenance, cooling, and physical security. “Space serves exclusively as an energy source, not an operational site,” Meta officials said. This approach is intended to reduce the complexity and cost of managing data center operations in orbit, as grid connection is significantly more cost-effective than running data centers in space.

To support the continuous energy demands of AI infrastructure, Meta has also partnered with energy storage firm Noon Energy to secure up to 1 gigawatt of power capacity and 100 gigawatt-hours of long-duration energy storage. This storage solution will enable the retention of excess renewable energy generated during peak production periods and stabilize power availability during extended low-generation intervals, company sources confirmed. Meta described this as a critical component to address the intermittent nature of traditional renewable energy systems.

Meta’s broader energy strategy includes contracts for more than 30 gigawatts of clean, renewable energy, encompassing geothermal projects and agreements for 7.7 gigawatts of nuclear capacity. The company reported that 100% of the electricity used in its owned and operated data centers is matched with clean renewable energy, underscoring its commitment to securing reliable, long-term energy supplies to support future AI expansion.

Despite these plans, both the space-based solar power technology and the associated long-duration storage systems remain in early development stages, according to industry experts and company disclosures. Challenges such as satellite launch costs, in-orbit maintenance procedures, and the economic viability of scaling the technology have yet to be fully resolved. No commercial operation of space-based solar power has been realized to date, and the timeline for widespread deployment depends on the successful achievement of several technical milestones outlined in the partnership agreement.

Meta’s announcement on April 28, 2026, marks one of the most ambitious attempts to integrate space-based solar power into data center energy infrastructure. The company’s pilot phase will begin with a 25-megawatt capacity expansion, with plans for full deployment contingent on the technology’s performance during initial demonstrations. Officials indicated that the project aims to complement existing renewable energy sources and storage solutions within Meta’s extensive energy portfolio.

Meta launched its Creator Fast Track program in early 2024 to prioritize Reels on Instagram and Facebook, aiming to attract established TikTok and YouTube creators, company officials said. The initiative sought to boost short-form video content on Meta’s platforms by offering incentives and streamlined support to high-profile influencers, sources confirmed.

The Creator Fast Track program, which Meta introduced in January 2024, offers selected creators enhanced monetization options, early access to new features, and dedicated support teams to accelerate their growth on Instagram and Facebook Reels, company officials said. The program is designed to compete directly with TikTok and YouTube by attracting influencers who have already built substantial followings on those platforms, according to sources familiar with Meta’s strategy. Participants in the program receive tailored guidance on content optimization and priority placement in Reels feeds, aiming to increase visibility and engagement.

In a February 2024 interview, Mosseri noted that Reels accounted for over 20% of Instagram’s total daily engagement, underscoring the platform’s shift toward short-form video content.

Meta executives, including Instagram head Adam Mosseri, have publicly emphasized the company’s commitment to expanding Reels as a core format. Internal documents reviewed by industry analysts show that Meta has allocated significant resources toward incentivizing creators to produce Reels, including bonus payments tied to video performance metrics such as views and shares.

The program’s launch follows Meta’s broader push to capture market share in the short-form video space, which has been dominated by TikTok since its global rise in 2018 and YouTube Shorts since 2020. Meta’s Q1 2024 earnings report highlighted Reels as a key driver of user engagement growth, with average daily time spent on Reels increasing by 15% compared to the previous quarter, according to financial disclosures. The company has also expanded its advertising offerings within Reels, enabling brands to reach audiences through immersive, vertical video ads, as detailed in a March 2024 Meta marketing briefing.

Sources confirmed that Meta’s recruitment efforts through Creator Fast Track include outreach to creators with follower counts exceeding 500,000 on TikTok and YouTube, offering them financial incentives and platform support to transition or diversify their content on Instagram and Facebook. While exact enrollment numbers have not been publicly released, industry observers estimate that several hundred creators have joined the program since its inception. Meta officials declined to provide specific participation figures but confirmed that the program remains selective and focused on high-impact creators.

Analysts tracking social media trends note that Meta’s approach reflects a broader industry shift toward short-form video formats, with Reels now constituting a growing percentage of content consumption across Meta’s family of apps. This aligns with Meta’s strategic goal to increase ad revenue through formats favored by younger demographics, who are more likely to engage with brief, dynamic videos. The company has also introduced new creator tools, such as enhanced editing features and analytics dashboards, to support content production and performance tracking, according to product updates released in February 2024.

Meta’s efforts to attract TikTok and YouTube creators come amid increasing competition for influencer talent and advertising dollars. TikTok reported in its Q4 2023 earnings a global user base exceeding 1.2 billion monthly active users, while YouTube Shorts surpassed 50 billion daily views worldwide in early 2024, according to data from market research firms. Meta’s pivot to prioritize Reels through initiatives like Creator Fast Track represents a strategic attempt to reclaim engagement and creator loyalty within this competitive landscape.

The company continues to refine the program based on creator feedback and platform performance metrics, with plans to expand incentives and feature access in the coming months, officials said. Meta’s broader content strategy remains focused on integrating Reels more deeply across Instagram and Facebook, aiming to create a seamless, engaging experience that appeals to both creators and audiences. The success of this approach will likely influence Meta’s positioning in the evolving short-form video market throughout 2024 and beyond.

Meta expanded its Manus AI platform Tuesday to integrate Instagram Creator Marketplace directly into its workflow tools for Facebook and Instagram. The update allows marketers to automate creator discovery and outreach using verified data from over 2 million creators across 18 markets, streamlining campaign management within the Manus environment, company officials said.

Marketers can automate the discovery and outreach process by filtering creators based on interests, audience demographics, and previous brand partnerships, all within the Manus environment, according to company statements. The supported markets include Argentina, Australia, Brazil, Canada, France, Germany, India, Indonesia, Israel, Japan, Mexico, the Netherlands, New Zealand, South Korea, Spain, Turkey, the United Kingdom, and the United States. However, the connector is currently unavailable in European Union countries, sources confirmed.

The beta version of the new connector links Instagram Creator Marketplace directly into Manus AI workflows, providing access to a verified database of more than 2 million creators across 18 markets, Meta officials said.

The integration enables marketers to shift from manual browsing of the Creator Marketplace to a prompt-driven AI workflow. Marketers describe their criteria in natural language, and Manus AI matches creators accordingly, officials said. Outreach workflows are integrated with third-party tools such as Canva and Google Suite, allowing for seamless campaign execution without leaving the Manus platform. Brands can also upload existing creator lists to identify lookalike creators using first-party data, streamlining the process of expanding influencer partnerships, according to Meta’s product documentation.

Creator data is sourced from Meta’s authenticated database, which includes verified performance metrics and audience analytics such as age, gender, and location. Manus AI evaluates the alignment between creators and their audiences to identify optimal marketing partnerships, officials explained. The platform provides engagement rates, true reach, and other authenticated insights to support confident creator selection decisions. This data-driven approach aims to reduce friction between content creation and distribution on Instagram, according to company representatives.

To access the Instagram Creator Marketplace connector, businesses must have a verified Meta business portfolio and an Instagram Business Account linked to a Facebook Page. Operations must be conducted within one of the supported markets to use the connector. Multiple ad accounts can be connected and managed simultaneously within Manus, and the connection to Meta Ads Manager is read-only to prevent unintended campaign modifications, sources said.

Meta acquired Manus AI, a startup specializing in AI-powered marketing workflows, in December 2025. The expansion of Manus AI connectors across Meta’s marketing stack, including Ads Manager and WhatsApp Business, represents a step toward a fully integrated AI-powered workspace for businesses, company officials stated. The integration is designed to reduce the need for marketers to switch between multiple apps and tools for different marketing tasks. Manus AI agents operate in the background, connecting various tools to streamline workflows, according to internal sources.

The Meta Ads Manager Connector, also in early beta, allows real-time advertising data queries within Manus. It supports automated report generation and campaign performance analysis using natural language inputs. Manus translates raw data into reports, presentations, dashboards, or infographics, providing visual overviews and insights that help identify gaps and opportunities in Meta Ads strategies, company materials show.

In addition to creator discovery and campaign management, the Instagram Connector enables direct content creation and publishing from the Manus workspace. Supported formats include posts, carousels, Stories, and Reels with captions. The platform retrieves engagement data such as reach, likes, comments, shares, and saves, and provides strategic recommendations for improving content performance based on engagement analysis, officials said. A professional Instagram profile is required to use the content creation and publishing features.

The rollout of these connectors is part of Meta’s broader effort to integrate AI across its marketing tools and to enhance workflow automation for brands and marketers. The company continues to develop and expand the capabilities of Manus AI as part of its marketing technology ecosystem.

The Consumer Federation of America filed a lawsuit Tuesday in the District of Columbia Superior Court against Meta, alleging the company profited from scam advertisements on Facebook and Instagram. According to the complaint, Meta misled users about its efforts to combat fraud while generating more than 10% of its revenue from high-risk ads, violating the District’s Consumer Protection Procedures Act.

The lawsuit, filed Tuesday in the District of Columbia Superior Court, contends that Meta violated the District’s Consumer Protection Procedures Act by allowing scam advertisements to proliferate on Facebook and Instagram while profiting from them. The Consumer Federation of America (CFA), a nonprofit advocacy group based in Washington, D.C., brought the suit on behalf of consumers harmed by the alleged deceptive practices. According to the complaint, Meta misled users about its efforts to combat fraud, downplaying the scale of scams and creating a false impression of safety on its platforms.

Internal documents cited in the lawsuit reveal that Meta anticipated generating more than 10% of its revenue in 2024 from high-risk ads, including scam advertisements, illegal gambling, and prohibited goods.

The CFA states that this amounts to approximately $16 billion in profits for the year, a figure that matches the reported losses Americans faced due to online scams in 2024. The complaint asserts that Meta charges higher rates to advertisers deemed higher risk rather than banning them outright, effectively prioritizing revenue over user safety.

The CFA alleges that Meta’s policies allow scam ads to persist despite numerous user reports and that the company routinely ignores these complaints. The lawsuit points to Meta’s practice of permitting up to 32 automated strikes for fraudulent activity before banning an advertiser, a threshold CFA says enables the continued spread of scams. According to the complaint, Meta’s terms of service and community standards exaggerate the company’s protective measures, promising “appropriate action” and removal of deceptive content that the CFA claims are not fulfilled in practice.

In response, Meta issued a statement calling the allegations a misrepresentation of its work and said it would vigorously defend itself in court. The company highlighted that it removed more than 159 million scam ads last year, with 92% taken down before any user reports. Meta also noted it disabled 10.9 million accounts linked to criminal scam operations on Facebook and Instagram. A spokesperson told AARP and CBS News that fighting scams is “bad for business” and that Meta aggressively combats fraudulent activity to protect people and businesses.

Ben Winters, director of AI and data privacy at the CFA, said in a statement that Meta prioritizes profit over user safety and knowingly inflates its bottom line at the expense of consumers. The CFA is seeking to recover the profits it alleges are “illegal” under the D.C. Consumer Protection Procedures Act on behalf of those harmed by scam ads on Meta’s platforms.

The lawsuit comes amid ongoing scrutiny of Meta’s handling of scam advertisements. A 2025 Wall Street Journal report revealed that employees had warned company leadership about the prevalence of scams but that revenue concerns prevented stronger enforcement actions. Additionally, a prior lawsuit against Meta was revived by the 9th U.S. Circuit Court of Appeals, which ruled that Section 230 immunity does not shield the company from claims related to breaches of its terms of service. That case, initially dismissed by a trial judge, was sent back for trial, signaling continued legal challenges over Meta’s responsibility for content on its platforms.

The CFA’s complaint echoes concerns raised by other watchdog groups and media investigations regarding the persistence of scam ads on social media. YouTube reporting confirmed that the CFA filed the complaint alleging Meta knowingly allowed scam advertisements to remain active to maintain revenue streams. The outcome of this latest lawsuit could have significant implications for how social media companies regulate advertising content and address consumer protection laws in the District of Columbia and beyond.

Meta enhanced its Threads platform with new creator tools and API updates in early 2026, including reply filters, richer publishing features, and a creator hub, officials said. The changes aimed to improve content management, third-party integrations, and analytics for creators and developers across the platform.

Meta’s Threads platform introduced a suite of API enhancements in early 2026 designed to bolster third-party integrations and streamline content management, officials said. The updates enable developers to support features such as ghost posts, cross-sharing to Instagram Stories, text attachments, GIFs, spoiler tags, reply approvals, and real-time webhooks for publishing or deleting content, according to Meta for Developers documentation and company statements. These improvements facilitate advanced publishing, discovery, and moderation workflows, officials noted, aligning with Meta’s broader platform changes targeting creators.

The rollout coincided with Threads reaching 300 million active users, a milestone Meta highlighted as expanding opportunities for creators to reach broader audiences.

In addition to API upgrades, Threads rolled out new reply filters on iOS devices, offering users options to sort replies by “Top,” “Recent,” “All,” and “Following,” according to the Threads changelog and official app update logs. The reply filtering features, initially available since November 2025, saw expanded iOS-specific deployment in early 2026, allowing users greater control over managing conversations in busy threads. The update logs confirm the availability of these sorting options, which aim to reduce clutter and improve user experience.

February 2026 saw the launch of a new creator hub on Threads, which provides tips, insights, and analytics to help creators grow their presence on the platform. The hub includes enhanced analytics capabilities and integrations with Spotify-hosted podcasts, featuring profile previews and dedicated episode details, sources said.

Threads is also experimenting with community-focused features, including topic-specific communities that users can join, along with a “Live Chat” function for community engagement. Moderators can set chat names, photos, start and end dates, and control participant access, according to platform testing reports and official update notes from January and February 2026. These features remain in development, with experimentation aimed at improving moderation and fostering real-time interaction within topic-based groups.

The platform’s post creation process was streamlined to include quick-access topics and communities directly below the composer, enhancing content categorization and discovery. A new “Dear Algo” feature, launched in February 2026, allows users to influence their algorithmic feeds, while automatic thread counts such as “1/3” help users navigate multi-part posts. Topic Tags, presented as pill-style labels similar to hashtags, were introduced to highlight trending conversations. Meta encourages users to edit cross-posted Instagram content by adding relevant topics, according to recent update briefs.

Search and discovery tools on Threads received notable improvements, including a “From: username” profile search feature for precise content discovery by creator, sources confirmed. The platform is testing additional search capabilities such as date range filters, expanded trending topics, and AI-powered summaries. Enhancements to the in-app browser on iOS and modifications to post link structures to include usernames were also implemented, supporting better niche conversation discovery and follower growth, according to platform reports.

On the monetization and engagement front, Meta officials, including Adam Mosseri, acknowledged efforts to address “rage bait” posts through increased awareness and resolution strategies. Active status indicators were introduced to display user logins, promoting real-time engagement opportunities. Creator payouts, which ceased in mid-2025, have shifted toward custom monetization systems planned for 2026. Features such as reposts, quotes, and trending topics remain central to audience-building strategies. The platform is also testing default feed options allowing users to choose between “Following,” “For You,” or custom feeds.

These updates continue Meta’s iterative approach to refining Threads as a creator-focused platform, with ongoing experiments in community engagement and content moderation. The company has not announced specific launch dates for all features but tracks many as part of its January and February 2026 update cycles. As Threads evolves, Meta aims to provide developers, creators, and brands with enhanced tools to manage content and grow their audiences across its social media ecosystem.

Meta integrated its Manus AI technology into Ads Manager, Instagram Creator Tools, and WhatsApp Business starting February 17, 2026, officials said. The move followed Meta’s $2 billion acquisition of Manus AI in December 2025 and aims to automate multi-step advertising and content tasks using natural language prompts and anonymized chat signals.

The integration of Manus AI into Meta’s Ads Manager allows advertisers to automate complex, multi-step tasks through natural language prompts, officials said. The AI also produces a “Meta Ads Manager Connector Capabilities Brief” summarizing account data and creates custom dashboards for visualizing ad performance; however, these dashboards provide read-only access and do not permit live campaign modifications, sources confirmed.

Starting February 17, 2026, users can access Manus AI via the Tools menu in Ads Manager, where it generates natural language performance reports from campaign data and responds to plain English queries about metrics and anomalies, according to Meta spokespersons.

Meta’s use of anonymized AI chat signals from WhatsApp, Messenger, and Instagram further enhances ad targeting models, officials said. These signals, which influence ad personalization for over 1 billion monthly Meta AI users, are collected only if WhatsApp accounts are linked to the Meta Accounts Center. This integration represents a direct monetization of Meta’s conversational AI investments, according to company statements. The use of anonymized chat data is one of three key initiatives Meta has launched for advertisers in early 2026.

On Instagram, Manus AI is integrated into Creator Tools in an early beta phase, enabling professional profiles to create and publish posts, carousels, Stories, and Reels with captions directly from the Manus workspace, sources said. The AI also retrieves engagement metrics such as reach, likes, comments, shares, and saves, and offers strategic recommendations to improve content performance. Additionally, the Manus-powered Meta Ads platform includes an Instagram Ads generator optimized for conversions, according to internal documents.

WhatsApp Business users also benefit from the integration, with AI conversations processed as intent signals for ad personalization, officials noted. These signals are included only if the WhatsApp account is linked to Facebook or Instagram through the Meta Accounts Center. This feature is particularly relevant in markets where WhatsApp serves as the primary communication platform. The WhatsApp signals contribute to the broader anonymized data set used across Meta’s platforms for ad targeting enhancements introduced in early 2026.

In March 2026, Meta launched the Small Business AI program under the leadership of Dina Powell McCormick. The initiative offers free AI-powered advertising tools to approximately 250 million small businesses using Meta platforms, company records show. The program includes AI business assistants, automated campaign creation, and budget optimization features designed to streamline advertising efforts for small enterprises. This program is identified as the third major move in Meta’s early 2026 advertising strategy.

Performance data indicates that campaigns using Meta’s Advantage+ feature achieve a 22% higher return on ad spend compared to manual management, according to internal analyses. The Manus Generative Ads Model (GEM) has increased ad conversions by 5% on Instagram and 3% on Facebook Feed, sources confirmed. These improvements impact more than 1 billion Meta AI users through the integration of chat signals and serve the extensive small business user base via AI tools.

Meta acquired Manus AI in December 2025 for approximately $2 billion, a deal confirmed by multiple sources including YouTube analysis channels. The acquisition enabled rapid embedding of Manus technology into Meta platforms, officials said. Manus AI is described as a general-purpose autonomous AI agent capable of managing multi-step tasks with minimal supervision.

The early beta integrations of Manus AI features are not yet available to all users, according to company representatives. Meta continues to expand the rollout of these AI tools across its platforms as part of its broader advertising automation strategy.

Meta announced Tuesday it expanded the Threads API to allow third-party apps to manage reply approvals and include text attachments in scheduled posts. According to company officials, the updates enable users to control which replies appear publicly and support enhanced content scheduling features like GIFs and spoiler tags.

The update introduces reply approvals, a feature that allows Threads users to moderate which replies become publicly visible before they appear to others, according to Meta officials. This capability, now accessible through the Threads API, enables third-party applications to manage pending replies and maintain conversation quality by filtering responses in real time. Meta described the feature as a tool for users to “set the tone of the conversation” with greater control over engagement, as reported by TechCrunch on Thursday.

The Threads Reply Management API requires specific permissions—`threads_basic`, `threads_manage_replies`, and `threads_read_replies`—and enforces a limit of 1,000 API-published replies per profile within a 24-hour period, according to Meta’s developer documentation.

Alongside reply approvals, the expanded API supports text attachments within scheduled posts, enhancing content creation options for social media managers using external platforms. These attachments can include GIFs and spoiler tags for text, images, and videos, allowing for richer and more diverse Threads content. Social Media Today highlighted that these additions facilitate integrated scheduling and posting capabilities, broadening the functionality of third-party tools.

Meta also extended developer options to pre-configure topic tags, set reply controls, and compose replies or quotes directly through the API. Developers can now embed public Threads posts on external websites without requiring an API access token, increasing the accessibility of Threads content beyond the app itself.

Further enhancements include real-time notifications for post publications, deletions, and reply management events, which improve the responsiveness of third-party applications managing Threads engagement. Analytics capabilities have been upgraded to allow searching public posts by media type or author username, and the follower threshold for profile discovery has been lowered from 1,000 to 100 followers, enabling broader access to third-party insights. Additional sorting options now allow filtering verified users or those with and without profile images, according to sources familiar with the update.

User controls have been refined with activity feed filters that now include views for replies from followed users or mentions, complementing existing filters for verified accounts, quotes, and reposts. Instagram head Adam Mosseri noted ongoing tests aimed at tuning the Threads algorithm based on topic interests, reflecting Meta’s continued efforts to optimize user experience across its platforms.

The API expansion supports Threads’ growth, which has reached approximately 400 million monthly active users, as reported by Global Dating Insights. These developments are designed to empower social media managers and developers by integrating posting, scheduling, moderation, and analytics tools into third-party platforms, according to company officials.

Meta’s announcement follows earlier API enhancements that introduced ghost posts and the ability to share Threads content to Instagram Stories via external applications. These updates collectively aim to strengthen the Threads ecosystem by enabling more robust third-party integrations, as outlined in Meta’s official release and corroborated by industry reports from TechCrunch and Social Media Today.

The company continues to refine its API offerings to accommodate the evolving needs of developers and users, positioning Threads as a more versatile platform within Meta’s social media portfolio.

Meta updated its Threads API on Tuesday to allow third-party tools to support ghost posts, spoiler tags, and cross-platform sharing, among other features. According to company officials, the enhancements aim to improve content discovery, moderation workflows, and developer integration for a broader range of users and applications.

The updated Threads API introduces several key features designed to enhance content creation and moderation through third-party tools, Meta officials said Tuesday. Among the new capabilities, ghost posts are now supported, allowing users to schedule posts that can be published without immediate visibility, according to Meta’s developer documentation. Scheduled posts also now accommodate text attachments and GIFs, expanding creative options for social media managers. Spoiler tags can be applied to text, images, and videos, providing users with greater control over content presentation. Additionally, cross-platform sharing enables direct posting of Threads content to Instagram Stories, a move aimed at broadening audience reach across Meta’s platforms.

The follower threshold for profile discovery has been lowered from 1,000 to 100 followers, facilitating broader account surfacing and making it easier for external tools to identify a larger pool of public profiles.

Meta representatives highlighted improvements in content discovery and profile visibility, noting that public post searches can now be filtered by media type or author username. Sorting options have been added to distinguish verified users or those with and without profile images, officials said. These enhancements aim to support research, listening, and creator discovery by providing richer data and more nuanced search capabilities.

Developer integration has also been strengthened with the introduction of a tokenless oEmbed API, which allows embedding of public Threads posts without requiring API access tokens, according to Meta’s API reference. Updated post intent parameters offer developers more options for integration, while topic tag pre-configuration enables users to set tags within external applications. Third-party tools can now manage reply controls and support quote post composition, expanding the range of interactions possible from outside the Threads platform.

Moderation workflows have been enhanced with the addition of reply approvals and pending reply management, providing better control over user interactions, sources confirmed. Real-time webhooks now notify developers of publishing and deletion events, allowing for more responsive moderation. For Threads ads, top-level reply moderation features have been introduced, including the ability to view, hide, and respond to replies. Access to these moderation tools requires ad account permissions, ensuring that only authorized users can manage content visibility and interactions.

On the advertising front, Meta has made app ads available globally on Threads through the Marketing API, officials said. Supported objectives include app installs and app event optimization, with Threads feed placement expanding campaign reach. Advertisers can reuse existing creative assets without needing platform-specific production, streamlining campaign management. Marketing API access for these features requires “ads_read” or “ads_management” permissions, according to Meta’s advertising guidelines.

Operational integration improvements aim to reduce friction between Threads and existing social media management stacks. Meta representatives noted that the update supports single-platform integration for planning, approvals, scheduling, reporting, and community management. Third-party customer relationship management (CRM) and social media management platforms are now supported, enabling Threads to function more naturally within established workflows. These changes are intended to enhance professional management accessibility by allowing deeper platform integration.

Meta reported that Threads currently has 400 million monthly active users, a user base that continues to grow and supports the expanded API functionality, company officials said. The strategic importance of search and discovery updates was emphasized as a factor in scaling the platform’s reach. Broader accessibility for brands is facilitated through professional management tools, while enhanced content engagement tracking is enabled by expanded analytics options included in the API update.

The Threads API update follows Meta’s broader efforts to unify its social media offerings and provide developers with more robust tools to manage content and advertising across platforms. The company continues to invest in features that support a wide range of users, from individual creators to large brands, aiming to integrate Threads more closely with existing marketing and operational ecosystems.

• Marking a major investment in Meta’s AI future, we are announcing two 24k GPU clusters. We are sharing details on the hardware, network, storage, design, performance, and software that help us extract high throughput and reliability for various AI workloads. We use this cluster design for Llama 3 training.
• We are strongly committed to open compute and open source. We built these clusters on top of Grand Teton, OpenRack, and PyTorch and continue to push open innovation across the industry.
• This announcement is one step in our ambitious infrastructure roadmap. By the end of 2024, we’re aiming to continue to grow our infrastructure build-out that will include 350,000 NVIDIA H100 GPUs as part of a portfolio that will feature compute power equivalent to nearly 600,000 H100s.

To lead in developing AI means leading investments in hardware infrastructure. Hardware infrastructure plays an important role in AI’s future. Today, we’re sharing details on two versions of our 24,576-GPU data center scale cluster at Meta. These clusters support our current and next generation AI models, including Llama 3, the successor to Llama 2, our publicly released LLM, as well as AI research and development across GenAI and other areas .

A peek into Meta’s large-scale AI clusters

Meta’s long-term vision is to build artificial general intelligence (AGI) that is open and built responsibly so that it can be widely available for everyone to benefit from. As we work towards AGI, we have also worked on scaling our clusters to power this ambition. The progress we make towards AGI creates new products, new AI features for our family of apps, and new AI-centric computing devices. 

While we’ve had a long history of building AI infrastructure, we first shared details on our AI Research SuperCluster (RSC), featuring 16,000 NVIDIA A100 GPUs, in 2022. RSC has accelerated our open and responsible AI research by helping us build our first generation of advanced AI models. It played and continues to play an important role in the development of Llama and Llama 2, as well as advanced AI models for applications ranging from computer vision, NLP, and speech recognition, to image generation, and even coding.

Under the hood

Our newer AI clusters build upon the successes and lessons learned from RSC. We focused on building end-to-end AI systems with a major emphasis on researcher and developer experience and productivity. The efficiency of the high-performance network fabrics within these clusters, some of the key storage decisions, combined with the 24,576 NVIDIA Tensor Core H100 GPUs in each, allow both cluster versions to support models larger and more complex than that could be supported in the RSC and pave the way for advancements in GenAI product development and AI research.

Network

At Meta, we handle hundreds of trillions of AI model executions per day. Delivering these services at a large scale requires a highly advanced and flexible infrastructure. Custom designing much of our own hardware, software, and network fabrics allows us to optimize the end-to-end experience for our AI researchers while ensuring our data centers operate efficiently. 

With this in mind, we built one cluster with a remote direct memory access (RDMA) over converged Ethernet (RoCE) network fabric solution based on the Arista 7800 with Wedge400 and Minipack2 OCP rack switches. The other cluster features an NVIDIA Quantum2 InfiniBand fabric. Both of these solutions interconnect 400 Gbps endpoints. With these two, we are able to assess the suitability and scalability of these different types of interconnect for large-scale training, giving us more insights that will help inform how we design and build even larger, scaled-up clusters in the future. Through careful co-design of the network, software, and model architectures, we have successfully used both RoCE and InfiniBand clusters for large, GenAI workloads (including our ongoing training of Llama 3 on our RoCE cluster) without any network bottlenecks.

Compute

Both clusters are built using Grand Teton, our in-house-designed, open GPU hardware platform that we’ve contributed to the Open Compute Project (OCP). Grand Teton builds on the many generations of AI systems that integrate power, control, compute, and fabric interfaces into a single chassis for better overall performance, signal integrity, and thermal performance. It provides rapid scalability and flexibility in a simplified design, allowing it to be quickly deployed into data center fleets and easily maintained and scaled. Combined with other in-house innovations like our Open Rack power and rack architecture, Grand Teton allows us to build new clusters in a way that is purpose-built for current and future applications at Meta.

We have been openly designing our GPU hardware platforms beginning with our Big Sur platform in 2015.

Storage

Storage plays an important role in AI training, and yet is one of the least talked-about aspects. As the GenAI training jobs become more multimodal over time, consuming large amounts of image, video, and text data, the need for data storage grows rapidly. The need to fit all that data storage into a performant, yet power-efficient footprint doesn’t go away though, which makes the problem more interesting.

Our storage deployment addresses the data and checkpointing needs of the AI clusters via a home-grown Linux Filesystem in Userspace (FUSE) API backed by a version of Meta’s ‘Tectonic’ distributed storage solution optimized for Flash media. This solution enables thousands of GPUs to save and load checkpoints in a synchronized fashion (a challenge for any storage solution) while also providing a flexible and high-throughput exabyte scale storage required for data loading.

We have also partnered with Hammerspace to co-develop and land a parallel network file system (NFS) deployment to meet the developer experience requirements for this AI cluster. Among other benefits, Hammerspace enables engineers to perform interactive debugging for jobs using thousands of GPUs as code changes are immediately accessible to all nodes within the environment. When paired together, the combination of our Tectonic distributed storage solution and Hammerspace enable fast iteration velocity without compromising on scale.     

The storage deployments in our GenAI clusters, both Tectonic- and Hammerspace-backed, are based on the YV3 Sierra Point server platform, upgraded with the latest high capacity E1.S SSD we can procure in the market today. Aside from the higher SSD capacity, the servers per rack was customized to achieve the right balance of throughput capacity per server, rack count reduction, and associated power efficiency. Utilizing the OCP servers as Lego-like building blocks, our storage layer is able to flexibly scale to future requirements in this cluster as well as in future, bigger AI clusters, while being fault-tolerant to day-to-day Infrastructure maintenance operations.

Performance

One of the principles we have in building our large-scale AI clusters is to maximize performance and ease of use simultaneously without compromising one for the other. This is an important principle in creating the best-in-class AI models. 

As we push the limits of AI systems, the best way we can test our ability to scale-up our designs is to simply build a system, optimize it, and actually test it (while simulators help, they only go so far). In this design journey, we compared the performance seen in our small clusters and with large clusters to see where our bottlenecks are. In the graph below, AllGather collective performance is shown (as normalized bandwidth on a 0-100 scale) when a large number of GPUs are communicating with each other at message sizes where roofline performance is expected. 

Our out-of-box performance for large clusters was initially poor and inconsistent, compared to optimized small cluster performance. To address this we made several changes to how our internal job scheduler schedules jobs with network topology awareness – this resulted in latency benefits and minimized the amount of traffic going to upper layers of the network. We also optimized our network routing strategy in combination with NVIDIA Collective Communications Library (NCCL) changes to achieve optimal network utilization. This helped push our large clusters to achieve great and expected performance just as our small clusters.

In the figure we see that small cluster performance (overall communication bandwidth and utilization) reaches 90%+ out of the box, but an unoptimized large cluster performance has very poor utilization, ranging from 10% to 90%. After we optimize the full system (software, network, etc.), we see large cluster performance return to the ideal 90%+ range.

In addition to software changes targeting our internal infrastructure, we worked closely with teams authoring training frameworks and models to adapt to our evolving infrastructure. For example, NVIDIA H100 GPUs open the possibility of leveraging new data types such as 8-bit floating point (FP8) for training. Fully utilizing larger clusters required investments in additional parallelization techniques and new storage solutions provided opportunities to highly optimize checkpointing across thousands of ranks to run in hundreds of milliseconds.

We also recognize debuggability as one of the major challenges in large-scale training. Identifying a problematic GPU that is stalling an entire training job becomes very difficult at a large scale. We’re building tools such as desync debug, or a distributed collective flight recorder, to expose the details of distributed training, and help identify issues in a much faster and easier way

Finally, we’re continuing to evolve PyTorch, the foundational AI framework powering our AI workloads, to make it ready for tens, or even hundreds, of thousands of GPU training. We have identified multiple bottlenecks for process group initialization, and reduced the startup time from sometimes hours down to minutes. 

Commitment to open AI innovation

Meta maintains its commitment to open innovation in AI software and hardware. We believe open-source hardware and software will always be a valuable tool to help the industry solve problems at large scale.

Today, we continue to support open hardware innovation as a founding member of OCP, where we make designs like Grand Teton and Open Rack available to the OCP community. We also continue to be the largest and primary contributor to PyTorch, the AI software framework that is powering a large chunk of the industry.

We also continue to be committed to open innovation in the AI research community. We’ve launched the Open Innovation AI Research Community, a partnership program for academic researchers to deepen our understanding of how to responsibly develop and share AI technologies – with a particular focus on LLMs.

An open approach to AI is not new for Meta. We’ve also launched the AI Alliance, a group of leading organizations across the AI industry focused on accelerating responsible innovation in AI within an open community. Our AI efforts are built on a philosophy of open science and cross-collaboration. An open ecosystem brings transparency, scrutiny, and trust to AI development and leads to innovations that everyone can benefit from that are built with safety and responsibility top of mind. 

The future of Meta’s AI infrastructure

These two AI training cluster designs are a part of our larger roadmap for the future of AI. By the end of 2024, we’re aiming to continue to grow our infrastructure build-out that will include 350,000 NVIDIA H100s as part of a portfolio that will feature compute power equivalent to nearly 600,000 H100s.

As we look to the future, we recognize that what worked yesterday or today may not be sufficient for tomorrow’s needs. That’s why we are constantly evaluating and improving every aspect of our infrastructure, from the physical and virtual layers to the software layer and beyond. Our goal is to create systems that are flexible and reliable to support the fast-evolving new models and research.