AI News on June 18, 2026

Research Breakthrough

Meituan LongCat Team Unveils WBench: The First Systematic Multi-Round Benchmark for Interactive Video World Models

Meituan's LongCat team has introduced and open-sourced WBench, a pioneering systematic multi-round evaluation benchmark designed specifically for interactive video world models. Positioned as a diagnostic 'CT scanner' for the AI industry, WBench is engineered to identify the precise technical bottlenecks encountered as world models transition from passive video generation to active, interactive environments. By providing a structured framework for multi-round assessment, the benchmark offers researchers a tool to pinpoint where current models fail during complex interactions. This release marks a significant step in standardizing the evaluation of dynamic AI systems, moving beyond traditional 'passive viewing' metrics to more rigorous, interaction-based performance analysis.

美团技术团队•June 18, 2026

Read AnalysisSource

Industry News

Meituan Showcases AI Innovations at ACL 2026: From Model Evaluation to Advanced Reasoning Paradigms

At the prestigious ACL 2026 conference, the Meituan technical team presented six groundbreaking papers that signal a shift toward a new generative paradigm in artificial intelligence. These research contributions span a diverse array of critical NLP and AI domains, including large-scale model evaluation, complex process reasoning, and the optimization of competition-level mathematical thinking. Additionally, the papers explore advancements in reinforcement learning and generative recommendation systems. By focusing on these specific technical directions, Meituan aims to enhance the reasoning capabilities and practical utility of AI models. This selection highlights Meituan's commitment to pushing the boundaries of computational linguistics and natural language processing, providing insights into how the industry can transition from simple generation to more sophisticated, optimized reasoning and recommendation frameworks.

美团技术团队•June 18, 2026

Read AnalysisSource

Open Source

LongCat-Video-Avatar 1.5 Open-Sourced: Meituan Advances Digital Human Video Models for Commercial-Grade Applications

Meituan's technical team has officially open-sourced LongCat-Video-Avatar 1.5, a significant upgrade in digital human video modeling. Transitioning from a state-of-the-art (SOTA) research model to a commercial-ready solution, version 1.5 introduces major improvements in lip-sync accuracy, physical realism, and long-form video stability. The model is designed to handle complex commercial environments, supporting multi-person interactions and offering high inference efficiency. By bridging the gap between experimental prototypes and real-world deployment, LongCat-Video-Avatar 1.5 enables the generation of high-quality, natural digital human content across diverse scenarios, moving the technology from the laboratory to the global stage.

美团技术团队•June 18, 2026

Read AnalysisSource

Industry News

Meituan LongCat Team Launches General 365 Benchmark: Gemini 3 Pro Leads with 62.8% Accuracy

The Meituan LongCat team has officially introduced General 365, a new benchmark designed to evaluate the reasoning capabilities of large language models. In a comprehensive assessment of 26 mainstream models, the results reveal a significant performance gap in the industry. Gemini 3 Pro, currently identified as the top-performing model, achieved an accuracy rate of 62.8%. However, the benchmark results highlight a broader challenge: the vast majority of tested models failed to reach the 60% accuracy threshold. This release establishes a new standard for measuring AI intelligence and underscores the current limitations of complex reasoning in even the most advanced AI systems.

美团技术团队•June 18, 2026

Read AnalysisSource

Industry News

Managing AI Coding Through Agent Evaluation: A Case Study of Refactoring 310,000 Lines of Code

The Meituan technical team has shared a comprehensive framework for managing AI-driven development, centered on the successful refactoring of 310,000 lines of code. As AI begins to generate over 90% of codebases, the team argues that the bottleneck has shifted from coding speed to the implementation of effective constraints. Without standardized management, AI risks magnifying system complexity and chaos. The team's approach utilizes 'Agent evaluation thinking' to transform refactoring from a high-cost, specialized project into a continuous daily activity. This is achieved through four key pillars: technical debt assessment, rule construction, standardized operating procedures (SOPs), and a Pre-PR (Pull Request) mechanism. This methodology ensures that AI-generated code remains aligned with system architecture and quality standards, providing a blueprint for sustainable AI-assisted software engineering.

美团技术团队•June 18, 2026

Read AnalysisSource

Research Breakthrough

LongCat-AudioDiT: Meituan's Breakthrough in Zero-Shot TTS Voice Cloning via Waveform Latent Space Diffusion

Meituan's LongCat team has unveiled LongCat-AudioDiT, a pioneering model designed to push the boundaries of zero-shot voice cloning. By abandoning traditional intermediate representations such as Mel-spectrograms, the model operates directly within the waveform latent space using a diffusion-based framework. This strategic shift is designed to eliminate cascade errors inherent in multi-stage data conversion, allowing the AI to learn the fundamental laws of sound directly. The result is a more streamlined and accurate Text-to-Speech (TTS) process that enhances the fidelity of voice cloning. This development represents a significant technical leap in the field of audio synthesis, focusing on architectural purity to enhance the authenticity of generated speech and overcoming long-standing technical bottlenecks in the industry.

美团技术团队•June 18, 2026

Read AnalysisSource

Open Source

LongCat-Flash-Prover: Meituan Open-Sources AI Model for Rigorous Mathematical Theorem Proving and Formalization

Meituan's technical team has officially open-sourced LongCat-Flash-Prover, a specialized AI model designed to bridge the gap between simple numerical calculation and rigorous mathematical theorem proving. While traditional AI models often focus on predicting the correct final answer, LongCat-Flash-Prover prioritizes the construction of strict logical chains. The model addresses a critical challenge in complex reasoning: the tendency for natural language ambiguity to undermine the integrity of a proof. By focusing on mathematical formalization, Meituan aims to transition AI capabilities from "guessing answers" to executing verifiable, rigorous proofs. This release marks a significant contribution to the open-source community, providing a tool specifically tuned for the high-precision requirements of formal logic and mathematical structures.

美团技术团队•June 18, 2026

Read AnalysisSource

Research Breakthrough

LARYBench Released: Defining the ImageNet for Embodied Action Representations and Measuring Generalization from Human Videos

Meituan Technical Team has officially released LARYBench (Latent Action Representation Yielding Benchmark), a systematic framework designed to evaluate and guide the learning of general latent action representations from large-scale visual data. The benchmark's findings represent a significant breakthrough in embodied AI, revealing that general vision models outperform specialized action expert models in both action generalization and control precision. Most notably, the research demonstrates that embodied action representations can emerge naturally from large-scale human video data. By establishing a standardized metric for action representation, LARYBench aims to serve as the 'ImageNet' for the field of embodied intelligence, providing a clear path for developing more versatile and precise robotic control systems based on universal visual foundations.

美团技术团队•June 18, 2026

Read AnalysisSource

Open Source

Meituan Unveils LongCat-Next: A Native Multimodal Model for Real-World AI Perception and Interaction

Meituan's technical team has officially announced the release and open-sourcing of LongCat-Next, a native multimodal model designed to bridge the gap between artificial intelligence and the physical world. By treating vision and speech as "native languages," LongCat-Next represents a significant shift toward AI systems that can perceive, understand, and act within real-world environments. Alongside the model, Meituan has open-sourced its discrete tokenizer, providing the developer community with the foundational tools necessary to build sophisticated, multi-sensory AI applications. This initiative underscores Meituan's commitment to advancing the field of physical-world AI through collaborative, open-source research and development.

美团技术团队•June 18, 2026

Read AnalysisSource

Industry News

Meituan BI Evolution: Implementing Metric Platforms and Analysis Engines for Enhanced Data Consistency

Meituan's technical team has unveiled a new generation of Business Intelligence (BI) architecture centered on a centralized Metric Platform. This strategic shift aims to resolve persistent issues found in traditional BI environments, such as "data caliber confusion" and poor query performance. By developing two core capabilities—Automatic Semantics and Enhanced Computing—Meituan has successfully addressed the limitations of personalized dataset-driven models. This new framework ensures that data definitions remain consistent across the organization while significantly optimizing the speed and efficiency of data analysis. The implementation marks a significant milestone in Meituan's journey toward a more robust and scalable data infrastructure, providing a blueprint for handling complex enterprise-level BI challenges.

美团技术团队•June 18, 2026

Read AnalysisSource

Industry News

Leaked OpenAI Financials Reveal Massive Revenue Growth Amidst Multi-Billion Dollar Losses and Rising R&D Costs

Leaked financial documents, audited and reviewed by independent journalists and the Financial Times, reveal OpenAI's financial trajectory as it prepares for a potential IPO. While the company's revenue surged from $3.7 billion in 2024 to $13.07 billion in 2025, its expenses have grown even faster. Research and development costs reached a staggering $19.18 billion in 2025, driven largely by model training and payments to Microsoft. Additionally, the cost of revenue and sales marketing expenses have seen significant increases. Although OpenAI's operating loss is shrinking relative to its revenue, the company remains billions of dollars away from profitability, highlighting the immense capital requirements of leading the generative AI sector and the significant costs associated with inference and scaling.

Hacker News•June 17, 2026

Read AnalysisSource

Industry News

Grok 4.1 Fast Dominates AI Battle Royale Experiment While Claude Sonnet 4.6 Prioritizes Cooperation Over Combat

In a groundbreaking experiment conducted by Jacky Liang of OpenRouter, 11 Large Language Models (LLMs) were placed in a 2D battle royale simulation to test their competitive capabilities. The results revealed a stark contrast in performance and behavior: xAI’s Grok 4.1 Fast emerged as the dominant victor, winning 43% of the matches (13 out of 30) at a highly efficient cost of $0.97 per win. Conversely, Anthropic’s Claude Sonnet 4.6, despite being a top-tier model, won only 5 games and cost 27 times more per win. The experiment highlighted significant behavioral differences, with Claude attempting to form alliances and socialize, while GPT 5.4 led in total kills but failed to secure the most victories. This study suggests that traditional benchmarks may fail to capture the nuanced behavioral traits essential for real-world AI agent deployment.

Hacker News•June 17, 2026

Read AnalysisSource

Product Launch

VSCO Launches Studio Pro to Challenge Adobe with High-End Features and $500 Annual Subscription

VSCO has officially entered the professional creative software market with the launch of Studio Pro, a new editing application designed to compete directly with Adobe. Initially released for iOS, the app is scheduled for a macOS debut later this year. Studio Pro introduces high-efficiency tools such as batch editing and a style-matching feature that allows users to replicate the aesthetic of a reference image. Alongside these technical additions, VSCO is introducing a premium subscription tier priced at $500 per year, signaling a significant shift toward the high-end professional market. By integrating these tools with VSCO Galleries, the company aims to provide a streamlined workflow for creators who require both advanced editing capabilities and a platform for professional image sharing.

The Verge•June 17, 2026

Read AnalysisSource

Industry News

Snap Stock Declines Following Debut of High-Priced Augmented Reality Smart Glasses

Snap Inc. has officially introduced its highly anticipated augmented reality (AR) smart glasses, but the market's response has been decidedly negative. Despite the long-awaited nature of this hardware debut, the company's stock price experienced a significant downturn immediately following the announcement. The primary driver for this investor skepticism appears to be the product's pricing, which has been characterized as "ridiculously expensive." This financial reaction suggests a disconnect between the company's hardware ambitions and market expectations regarding consumer accessibility and value. The debut, which was expected to be a milestone for the company, has instead resulted in a decline in shareholder confidence, highlighting the risks associated with Snap's transition into high-end wearable technology.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

NEA Partner Tiffany Luck Highlights Enterprise Challenges in Determining Artificial Intelligence Return on Investment

The initial wave of AI enthusiasm, characterized by the 'tokenmaxxing' trend, is facing a reality check as enterprises struggle to define clear Return on Investment (ROI). According to Tiffany Luck of NEA, the period of encouraging employees to maximize AI usage has led to significant budgetary strains. High-profile cases include Uber reportedly exhausting its entire annual AI budget within a few months, while other organizations have begun cutting licenses for tools like Claude. Even tech giants like Meta are pivoting, recently dismantling internal AI usage leaderboards. This shift signals a transition from experimental adoption to a more disciplined financial approach, as businesses move to reconcile the high costs of AI tokens with tangible business outcomes.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

Anthropic Makes History as the First AI Startup to Join the Frontier Carbon Removal Coalition

Anthropic has officially joined the Frontier coalition, marking a significant milestone as the first artificial intelligence startup to participate in this climate-focused initiative. The announcement coincides with Frontier securing an additional $915 million in pledges, which are specifically designated to fund various carbon removal projects. This move highlights a growing trend of high-growth AI companies taking an active role in environmental sustainability. By joining the coalition, Anthropic aligns itself with a major financial commitment aimed at accelerating technologies that remove carbon from the atmosphere. The $915 million in new funding represents a substantial boost for the carbon removal sector, emphasizing the increasing corporate interest in addressing climate change through innovative technological solutions and strategic partnerships.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

The Evolution of Social Media: How User-Controlled Algorithms are Transforming Threads, Instagram, and TikTok Feeds

The social media landscape is undergoing a fundamental shift as major platforms move toward a model of user-centric curation. According to recent industry reports, platforms including Threads, Instagram, and TikTok are introducing innovative tools that empower users to directly influence the algorithms responsible for their content recommendations. This transition marks the 'next evolution' of social media, moving away from a passive consumption model where platforms unilaterally dictate content flow. By providing enhanced customization options, these services aim to offer a more personalized and intentional user experience. This strategic move reflects a broader industry trend toward transparency and user agency, allowing individuals to shape their digital environments and have a direct say in the logic that governs their social media feeds.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

NEA’s Tiffany Luck on the AI ROI Reckoning: From Tokenmaxxing to Budgetary Discipline

The AI industry is currently navigating a significant transition from the aggressive 'tokenmaxxing' trend to a period of strict financial scrutiny, according to NEA's Tiffany Luck. This shift, characterized as an 'ROI reckoning,' comes as major tech entities face the reality of soaring AI costs. Notable examples include Uber reportedly exhausting its annual AI budget in just a few months and Meta discontinuing its internal AI leaderboard. As companies scale back on resources like Claude licenses, the focus is shifting toward the sustainability of AI investments. This analysis explores the implications of these budgetary tensions on the future of AI-driven IPOs, the development of personal agents, and the broader enterprise strategy for integrating artificial intelligence.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

World Model Maker Odyssey Reaches $1.45 Billion Valuation with Strategic Backing from Amazon and Major Investors

Odyssey, a pioneering developer in the field of world models, has officially reached a valuation of $1.45 billion following a significant funding round. The investment features prominent backing from industry leader Amazon and other high-profile names, signaling a major shift in the artificial intelligence landscape. As the industry begins to look beyond the era of Large Language Models (LLMs), world models are being hailed as the next frontier in AI development. This valuation not only cements Odyssey's position as a critical startup to watch but also highlights the growing institutional confidence in technologies that aim to move past the limitations of current generative AI. The involvement of major players like Amazon underscores the strategic importance of world models in the future of the global AI ecosystem.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

Pew Research Reveals Two-Thirds of Americans Believe AI is Advancing Too Quickly Despite Surging Chatbot Adoption

A comprehensive new study from Pew Research highlights a growing tension in the American public's relationship with artificial intelligence. While chatbot usage has climbed significantly to 49% of the population—up from just 33% in 2024—a substantial 63% of Americans express concern that the technology is evolving at an excessive pace. The data specifically points to the meteoric rise of ChatGPT, which has seen its user base double since 2023, now reaching 44% of the public. This analysis explores the disconnect between the rapid integration of AI into daily life and the increasing collective anxiety regarding the speed of innovation, suggesting that while utility is driving adoption, the industry faces a significant challenge in aligning its progress with public comfort levels and societal readiness.

The Verge•June 17, 2026

Read AnalysisSource

Industry News

New Pew Research Study Reveals Only 16 Percent of Americans View AI Impact as Positive for Society

A recent report from Pew Research highlights a significant divide between financial markets and public sentiment regarding Artificial Intelligence. While Wall Street continues to show strong enthusiasm for AI developments, only 16 percent of Americans believe the technology will have a positive impact on society. This stark contrast suggests a growing skepticism among the general population despite the rapid integration of AI into various sectors. The study underscores a disconnect between the economic valuation of AI and the social perception of its long-term consequences, indicating that the majority of the U.S. population remains unconvinced of the technology's benefits. The findings suggest that the industry's financial success is not currently translating into public confidence or perceived social value.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Industry News

Amazon Echo Dot Max Hits All-Time Low Price of $64.99 in Early Prime Day Sale

Amazon has officially launched its early Prime Day promotions, featuring record-breaking discounts on its first-party hardware. The standout offer is the Echo Dot Max, which is currently available for $64.99, marking a $35 reduction from its standard retail price. This price point represents a new all-time low for the device. The sale comes just one week ahead of the official Prime Day event, with several other Echo speakers also seeing significant price cuts. Industry analysts view these early deals as a strategic move by Amazon to dominate the smart home market before the main shopping holiday begins next week.

The Verge•June 17, 2026

Read AnalysisSource

Industry News

Google Launches New $99 Gemini-Powered Smart Speaker to Replace Traditional Google Assistant Commands

Google is making a significant strategic move to revitalize the smart home market by introducing a new $99.99 Google Home Speaker powered by Gemini generative AI. This new hardware marks a departure from the legacy Google Assistant era, which was characterized by rigid and specific voice commands. By integrating Gemini, Google aims to provide a more natural and conversational user experience, allowing the device to understand and respond to fluid dialogue rather than just pre-programmed triggers. This launch represents Google's core bet that generative AI can reinvent the utility and appeal of smart speakers, transitioning them from simple command-execution tools into sophisticated conversational companions for the home environment.

TechCrunch AI•June 17, 2026

Read AnalysisSource

Product Launch

Snap Unveils $2,195 Specs: Evan Spiegel’s 12-Year Vision to Humanize Computing and Transform Wearable Technology

Snap has officially debuted its latest hardware innovation, the new Specs, priced at a premium $2,195. In a recent interview with CNBC, Snap CEO Evan Spiegel revealed that the device is the result of more than 12 years of internal development. Spiegel positioned the high-end glasses as a strategic attempt to "bring computing into the world" and "make it more human." This launch represents a significant milestone for the company, moving beyond its social media roots to offer a sophisticated device designed to assist users in their daily lives. The high price point and the decade-long development cycle underscore Snap's commitment to redefining how technology integrates with the physical environment, focusing on a more natural and human-centric computing experience.

The Verge•June 17, 2026

Read AnalysisSource