Llama 4 70B
Llama 4 70B is Meta’s mid-size open-weight model, widely used in production inference deployments and as a base for fine-tuning.
Released
April 5, 2025
Type
multimodal
License
open-weight
Context
1,000,000 tokens
Capabilities
Architecture
Parameters: 70B
Links
Llama 4 70B in the news
Decrypt · Sep 16, 2026
Meta May Have Found a Fix for Its 'Pervert Glasses'
Meta is reportedly developing a version of its smart glasses that would exclude cameras, a move aimed at addressing substantial privacy objections to AI-powered wearable technology. This potential iteration of the smart glasses seeks to mitigate concerns that have arisen due to the integrated cameras in previous models, which raised privacy issues for both users and those around them. The absence of a camera would fundamentally alter the functionality and perception of the device, shifting its focus away from visual recording and potentially towards other sensory inputs or data processing capabilities. Smart glasses, in general, represent a burgeoning category of wearable technology that integrates computing power and internet connectivity into eyewear. Companies like Meta, Google, and Snap have been investing in this space, envisioning a future where digital information and augmented reality experiences are seamlessly overlaid onto the user's view of the real world. However, the integration of cameras in these devices has consistently been a focal point of privacy debates. Critics argue that the always-on recording capability poses a threat to personal privacy, enabling covert surveillance and the unauthorized collection of images and videos of individuals without their consent. This has led to public apprehension and, in some instances, regulatory scrutiny. Meta's previous attempts at smart glasses, such as the Ray-Ban Stories collaboration, included cameras and faced criticism regarding privacy. While these devices offered features like photo and video capture, along with audio playback and voice commands, the presence of a visible LED indicator light was intended to signal when the camera was active. Despite these measures, the inherent privacy implications of a camera worn on the face remained a significant hurdle for widespread adoption and public acceptance. The development of a camera-less version suggests Meta is actively seeking to overcome these barriers by reimagining the core hardware of its wearable offerings. The potential benefits of a camera-less smart glasses model could include broader appeal to consumers who are hesitant about the privacy risks associated with camera-equipped wearables. Such a device might focus on functionalities like audio assistance, navigation, real-time translation, or displaying notifications and contextual information without the ability to capture visual data. This strategic pivot could allow Meta to tap into a market segment that has been deterred by privacy concerns, while still offering a connected and intelligent eyewear experience. The company's ongoing research and development in artificial intelligence and augmented reality could be leveraged to enhance the capabilities of a camera-free device, making it a compelling alternative in the evolving landscape of wearable technology.
TechCrunch · Sep 16, 2026
After accusations of selling ‘perv glasses,’ Meta prepares to sell a pair without a camera
Meta is preparing to release a new iteration of its smart glasses that will feature a camera-free design, a strategic move intended to mitigate ongoing privacy concerns and accusations of misuse that have plagued the product. The company has faced scrutiny, with critics labeling the existing smart glasses as "pervert glasses" due to potential privacy violations and the ability to record individuals without their consent. This new model aims to re-establish user trust and broaden the appeal of its wearable technology by removing the controversial recording capability. The original Ray-Ban Stories smart glasses, a collaboration between Meta and EssilorLuxottica, were launched in September 2021. These glasses integrated a camera, microphone, and speakers, allowing users to capture photos and videos, make calls, and listen to music or podcasts. However, the presence of a camera immediately raised privacy alarms, leading to widespread debate about surveillance and consent. While Meta implemented an LED light to indicate when the camera was active, this measure was not universally perceived as sufficient to prevent potential misuse. Meta's decision to remove the camera from the next generation of smart glasses signals a significant pivot in its wearable strategy. The company has invested heavily in the metaverse and believes that smart glasses are a crucial component of its long-term vision for augmented reality and seamless digital integration. By addressing the most significant privacy objection, Meta hopes to overcome a major hurdle that has limited the adoption and public acceptance of its smart glasses. The new product is expected to retain other smart functionalities, such as audio playback and voice assistant integration, while focusing on a more privacy-conscious user experience. The competitive landscape for smart glasses is evolving, with several tech giants exploring similar avenues. However, Meta's prominent position in social media and its ambitious metaverse goals make its smart glasses a key product in its ecosystem. The "pervert glasses" label, though harsh, reflects a genuine public apprehension that Meta must now actively counter. The camera-free approach is a direct response to these criticisms, aiming to demonstrate Meta's commitment to user privacy and responsible technology development. This move could potentially pave the way for wider adoption by users who were previously hesitant due to privacy concerns, while still offering a connected eyewear experience.
GSMArena · Sep 16, 2026
Meta is working on a pair of smart glasses without a camera, new report claims
Meta is reportedly developing a new iteration of its smart glasses that will intentionally exclude cameras, according to a recent report. This strategic decision aims to address privacy concerns that have been associated with wearable technology, particularly devices that can capture video or images without explicit user consent. The move suggests a potential shift in Meta's approach to augmented reality hardware, prioritizing user comfort and social acceptance over advanced visual recording capabilities in this particular product line. While specific details about the new smart glasses remain scarce, the absence of a camera implies a focus on other functionalities. These could include audio playback, real-time information display through a heads-up display (HUD), or integration with voice assistants. Meta has been actively exploring the potential of smart glasses as a platform for its metaverse ambitions, envisioning a future where digital information and interactions are seamlessly overlaid onto the physical world. Previous iterations of Meta's smart glasses, developed in partnership with Ray-Ban, have included cameras, which have drawn scrutiny from privacy advocates and the public. The decision to remove cameras from this upcoming model could be a direct response to this feedback and a broader trend towards more privacy-respecting consumer electronics. Companies across the tech industry are increasingly aware of the need to build trust with consumers regarding data collection and usage. By launching a camera-free version, Meta may be attempting to differentiate its product and appeal to a wider audience that might be hesitant to adopt camera-equipped smart glasses due to privacy fears. This could also pave the way for future devices that incorporate cameras with more robust privacy controls or for specific, opt-in use cases. Meta's ongoing investment in augmented and virtual reality technologies, including its Reality Labs division, underscores the company's long-term commitment to developing the metaverse. Smart glasses are seen as a crucial component for delivering immersive and interactive experiences. The development of a camera-free model indicates a willingness to adapt its hardware strategy based on market reception and evolving societal expectations around privacy. The success of this new design will likely depend on its ability to offer compelling features and a seamless user experience without compromising on its privacy-centric positioning.
Fast Company · Sep 16, 2026
Microsoft CEO Satya Nadella: AI companies should serve ‘humanity first’
Microsoft CEO Satya Nadella has articulated a strong stance on the ethical development of artificial intelligence, urging companies to "do what it takes to build stuff that serves humanity first and is in human control." Speaking at the All-In Summit in Los Angeles on Tuesday, Nadella underscored the critical need for comprehensive testing of AI technologies, advocating for the inclusion of outside testers to ensure robust safety and reliability. This call for caution comes as the artificial intelligence landscape experiences rapid advancement, prompting a broader industry discussion on safety protocols and the potential risks associated with increasingly powerful AI systems. Nadella's remarks followed a warning he issued just a day prior, cautioning developers against the premature release of AI products. He emphasized that "as the stakes get higher, one should take all the time they need," and that failure to do so could lead to a loss of "permission to operate." This perspective aligns with a growing chorus of concern from prominent figures within the tech industry. The debate has been amplified by the recent resignation of researcher Jacob Coxon from Anthropic, a leading AI safety and research company. Coxon publicly accused major AI laboratories of "gambling with our lives," a sentiment that resonated with Anthropic CEO Dario Amodei. Amodei subsequently published an essay calling for an industry-wide slowdown in the development of "frontier AI" – the most advanced and potentially transformative AI models. This call for a pause or deceleration in development was also supported by OpenAI CEO Sam Altman, a key figure in the AI space, and other senior tech executives. However, not all industry leaders share the same urgency for a slowdown. Nvidia CEO Jensen Huang and Meta CEO Mark Zuckerberg have presented more market-oriented viewpoints. They argue that the inherent liabilities and reputational risks associated with malfunctioning or unsafe AI products already provide sufficient incentives for companies to self-regulate and ensure product safety. Huang, for instance, stated that companies should refrain from releasing products or services if they lack confidence in their functionality, capability, or safety. Nadella echoed this sentiment, suggesting that many AI-related risks can be effectively managed through established engineering best practices and common-sense quality control processes. He drew an analogy to dealing with critical "showstopper bugs" in traditional software development, where such issues necessitate halting the entire development process until resolved. Despite these parallels to existing engineering challenges, Nadella acknowledged that AI introduces unique and novel risks. He specifically highlighted the importance of transparency for businesses regarding how AI systems utilize their data. "I want my privacy. I want to be able to embed my knowledge in a set of weights I control," Nadella stated, emphasizing the need to protect intellectual property and prevent its leakage. Furthermore, he identified autonomous AI agents as a new and significant category of "insider risk," envisioning scenarios where such agents could pose unforeseen challenges.
TechCrunch · Sep 16, 2026
Threads’ new features let podcasters promote shows and reach listeners
Meta's Threads platform is enhancing its capabilities for podcasters by rolling out a suite of new features designed to facilitate show promotion and audience engagement. These additions aim to position Threads as a more robust hub for podcast discovery and discussion, directly competing with platforms like X (formerly Twitter). The new tools include dedicated profile cards for podcasters, which will allow them to showcase their shows directly on their Threads profiles. Additionally, users will be able to share direct episode links, making it easier for listeners to access specific content. Transcripts will also be integrated, improving accessibility and searchability for podcast episodes shared on the platform. Guest tagging functionality is another key feature, enabling creators to credit and link to guests appearing on their shows, thereby expanding reach through collaborative promotion. To further support creators, Threads is introducing posting reminders, helping podcasters maintain a consistent presence and engage their audience more effectively. Audience insights will also be provided, offering creators data on listener engagement and reach within the Threads ecosystem. These developments signal Meta's strategic intent to capture a significant share of the audio content promotion market. By integrating these podcast-specific features, Threads seeks to attract a new demographic of creators and consumers, fostering a more dynamic and interactive environment for audio content. The platform's move into this space reflects a broader trend of social media networks seeking to diversify their content offerings and user engagement strategies beyond traditional text and image-based posts. The introduction of these tools is expected to benefit independent podcasters and established networks alike, providing them with new avenues to connect with potential listeners and grow their subscriber base. The emphasis on direct linking and discoverability aims to streamline the user experience for both creators and consumers, making it simpler to find and share podcast episodes. Meta's investment in these features underscores the growing importance of audio content in the digital media landscape and its commitment to evolving Threads into a comprehensive social networking and content discovery platform.