Blog
SeedRealtime Audio-Visual Full-Duplex LLM Released: Toward Omni-Modal Natural Interaction
As a native audio-visual full-duplex LLM, SeedRealtime can jointly understand audio, visual, and temporal information, accurately identify the interaction target and user intent, delivering a brand-new "watch, listen, and speak" experience.
One-take Creation, Flexible Referencing: Introducing Seedance 2.5
30-second long-form storytelling with fully upgraded multimodal referencing and editing capabilities.
From Speech to Audio Creation | Introducing the Seed Audio 1.0 Audio Creation Model
A unified framework for joint modeling of voice, sound effects, ambient sound, and other audio elements.
Beyond Generation, It Understands Design | Introducing Seedream 5.0 Pro
A multimodal image generation model that features advanced reasoning and efficient content creation.
EdgeBench: Measuring Real-World Environment Learning and Discovering a New Scaling Law
Agents exhibit a clear log-sigmoid pattern in environment learning.
Publications
Jul 6, 2026EdgeBench: Unveiling Scaling Laws of Learning from Real-World EnvironmentsJun 30, 2026Seed2.0 Model Card: Towards Intelligence Frontier for Real-World ComplexityMay 29, 2026Task-Focused Memorization for Multimodal AgentsMay 16, 2026Charon: A Unified and Fine-Grained Simulator for Large-Scale LLM Training and InferenceApr 26, 2026MegaScale-Omni: A Hyper-Scale, Workload-Resilient System for MultiModal LLM Training in ProductionApr 22, 2026Seed3D 2.0: Advancing High-Fidelity Simulation-Ready 3D Content GenerationApr 15, 2026Seedance 2.0: Advancing Video Generation for World ComplexityApr 11, 2026Protenix-v2: Broadening the Reach of Structure Prediction and Biomolecular DesignJan 27, 2026Post-LayerNorm Is Back: Stable, ExpressivE, and DeepDec 2, 2025GR-RL: Going Dexterous and Precise for Long-Horizon Robotic Manipulation