⭐Explore UNO-Bench: A Unified Omni-Modal Benchmark 🚀 Unified Framework: Efficient uni/omni-modal understanding evaluation 📊 Comprehensive Evaluation: 44 task types across 5 modality combinations; Perception & Reasoning coverage 🗒️ High-Quality Dataset: human centric construction with 98% cross-modality solvable ⚡ Efficiency: +90% faster datasets via auto-compression at 98% consistency (18 benchmarks) 🔥 Real-World Relevance: Authentic Chinese contexts & Multi-Step Open-Ended Questions for reliable assessment 🤖 Universal scoring model: supporting 6-type questions; 95% OOD model accuracys 🧠 Key Finding: omni-modal capability exhibits Power-law synergy (emergent ability) in strong models 🌍 We’re now focused on Chinese scenarios — and actively seeking partners to co-build English & multilingual versions! Let’s expand global impact together. 🤗 Open-Source: 【Hugging Face】github.com/meituan-longcat/U… 【GitHub】huggingface.co/datasets/meit… 【Paper】github.com/meituan-longcat/U…

🔥 LongCat-Flash-Omni: Multimodal + Low-Latency 🏆 Leading Performance among Open-Source Omni-modal Models ☎️ Real-time Spoken Interaction: Millisecond-level E2E latency 🕒 128K context + Supports > 8min real-time AV interaction 🎥 Multimodal I/O: Arbitrary Combination of Text/Image/Audio/Video Input → Text/Speech Output (w/ LongCat-Audio-Codec) ⚙ ScMoE architecture on LongCat-Flash: 560B Parameters, 27B Active 🧠 Training: Novel Early-Fusion Omni-modal training paradigm -> No Single Modality Left Behind 🚀 Efficient Infrastructure: With optimized modality-decoupled parallel training, Omni sustains >90% throughput of text-only training efficiency 🤗 Model open-sourced: 【Hugging Face】huggingface.co/meituan-longc… 【GitHub】github.com/meituan-longcat/L… 📱 LongCat APP is here—available for both iOS and Android! Scan the QR code to quickly try its awesome voice interaction features! 💻For the PC experience, you can click on LongCat.AI for a free trial!

🚀 LongCat-Video Now Open-Source: Text/Image-to-Video + Video Continuation in One Model 🏆 Text/Image-to-Video Performance Hits Open-Source SOTA 🎬 Minutes-Long High-Quality Videos: No Color Drift/Quality Loss (Industry-Standout) ⚙ 13.6B Params | Strong Open-Source DiT-Based Unified Multitask Video Base Model ⚡ C2F Pipeline + Block Sparse Attention: 720p/30fps Video in Minutes 🤗 Open-Source Links: GitHub：github.com/meituan-longcat/L… Hugging Face：huggingface.co/meituan-longc… Project Page：meituan-longcat.github.io/Lo…

🔥 Meituan LongCat Team introduces VitaBench! An Agent Benchmark for Complex Real-Life Scenarios VitaBench focuses on three high-frequency real-life domains: Food Delivery, Restaurants, and Travel Services, systematically and quantitively measuring agents' ability in Reasoning, Tool Use, and Adaptive Interaction. We believe that our data construction procedure shares some valuable insight about controllably building up high-quality & difficult agentic training tasks. Across 20+ top models, success rates remain low: cross-scenario Avg@4: 30% single-scenario Avg@4: 48.3% cross-scenario Pass@4(≥1 success in 4 tries) ≈ 60% cross-scenario Pass⁴(4/4 successes) ≈ 0% Agents show potential but lack the reliability needed for production. Even top reasoning models struggle with stability. 📢 Fully open-sourced: 🌐 Project & Leaderboards: vitabench.github.io 💻 Code: github.com/meituan-longcat/v… 🤗 Dataset: huggingface.co/datasets/meit…

🎉 LongCat-Audio-Codec is officially OPEN SOURCED! 🚀 -an audio codec solution optimized specifically for Speech LLMs. Key Breakthroughs: 1. Dual Tokens: Semantic and Acoustic Tokens are extracted in parallel at a low frame rate (16.7Hz / 60ms).This ensures both efficient modeling and full information integrity. 2. Ultra-Efficiency: LongCat-Audio-Codec maintains high intelligibility even at an extremely low bitrate, such as 0.43 kbps. 3. Real-Time Ready: Features a low-latency streaming decoder architecture. Latency is controlled to the hundred-millisecond level for real-time interaction. The integration of super-resolution in the decoder further enhances audio quality without extra models! This solution lowers technical barriers and optimizes resource efficiency for mobile/embedded Speech LLM deployment. 🔗 Code: Github: github.com/meituan-longcat/L… Huggingface: huggingface.co/meituan-longc…

🎉LongCat's Latest Updates: 1️⃣ API Access: LongCat-Flash-Thinking is now available via API (Claude Code integration compatible) 2️⃣ Daily Free Quota: Increased from 100,000 → 500,000 tokens/day 3️⃣ Apply for More: Submit request on "Usage Info" page for 5M tokens/day after approval 4️⃣ New Docs: Change Log & Claude Code Configuration 🔄 Updated: API Docs, Quick Start Guide, FAQ Explore the speed & power of LongCat with ease ⚡ 🔗 LongCat API Platform: longcat.chat/platform/usage 🤗 HuggingFace: huggingface.co/meituan-longc… 💻 GitHub: github.com/meituan-longcat/L…

🚀 LongCat-Flash-Thinking: Smarter reasoning, leaner costs! 🏆 Performance: SOTA open-source models on Logic/Math/Coding/Agent tasks 📊 Efficiency: 64.5% fewer tokens to hit top-tier accuracy on AIME25 with native tool use, agent-friendly ⚙️ Infrastructure: Async RL achieves a 3x speedup over Sync frameworks 🔗Model: huggingface.co/meituan-longc… 💻 Try Now: longcat.ai

🔥 LongCat API Platform is LIVE! 🆓 Developers: Claim your 100,000 FREE tokens per day and leverage our cutting-edge LLM capabilities. 🛠️ LongCat API Platform is compatible with two leading API formats - choose OpenAI API or Anthropic API based on your needs! Build, test, and deploy AI solutions with unmatched efficiency. 🚀 Start building today: longcat.chat/platform

🎉 Congrats to Meituan LongCat team on launching LongCat-Flash-Chat — a 560B MoE model now open-sourced! Powered by SGLang inference acceleration, it achieves high efficiency and strong benchmark results. Details in the blog 👉 lmsys.org/blog/2025-09-01-sg…

🚀 LongCat-Flash-Chat Launches! ▫️ 560B Total Params | 18.6B-31.3B Dynamic Activation ▫️ Trained on 20T Tokens | 100+ tokens/sec Inference ▫️ High Performance: TerminalBench 39.5 | τ²-Bench 67.7 🔗 Model: huggingface.co/meituan-longc… 💻 Try Now: longcat.ai