Pandaily
What it covers
Latest stories it covered
China reports AI inference data surpasses training data for first time
China says the amount of AI “inference” activity generated by models has overtaken the amount of data used to train them for the first time, signaling wider adoption of AI systems...
Meituan releases LongCat-2.5, a multimodal AI model for long-horizon coding tasks
Meituan has launched LongCat-2.5, a native multimodal AI model aimed at extended coding and “long-horizon” agentic tasks. The release, dated November 2, adds improved visual unders...
Iluvatar CoreX unveils Tiangai 300 GPU, claiming performance comparable to Nvidia H200
Chinese AI chipmaker Iluvatar CoreX has presented its new Tiangai 300 GPU and says it can match or exceed Nvidia’s H200 for real-world AI workloads. The company positioned the anno...
Apex Intelligence releases Apex-35B-V1.0, a general-purpose AI model built with automated research
Apex Intelligence announces the release of Apex-35B-V1.0, described as its first general-purpose artificial intelligence model. The company says the system is built using automated...
DeepSeek V5 model identifiers appear in Harness code repository tests
Multiple outlets report that identifiers for a future DeepSeek V5 model appear in code related to the Harness platform on GitHub. TechNode says three model strings—deepseek-v5, dee...
Baidu releases Unlimited OCR 2 to parse very long documents in one pass
Baidu releases “Unlimited OCR 2,” a new open-source OCR model designed to parse documents of very great length in a single pass. The system is built to keep decoding cost nearly co...
ByteDance launches AI studio for short-drama production using three in-house models
ByteDance launches a short-drama production studio powered by AI, aimed at covering multiple stages of making short dramas. The studio, called Dreamina (connected to ByteDance’s Ji...
Tencent open-sources Hy-SimulMT for real-time Chinese-English speech interpretation
Tencent has released Hy-SimulMT, an open-source AI model designed for simultaneous Chinese-English interpretation. The system continuously processes incoming speech and produces tr...
Kanzhun’s Nanbeige team open-sources Nanbeige4.3-33B, a 33B agentic coding model
Nanbeige, a coding and office-work model from Kanzhun’s Nanbeige team, releases Nanbeige4.3-33B as an open model on Hugging Face. The release positions the 33B model as a “teacher”...
DeepSeek and MiniMax get approval to use AI models overseas, state media says
China’s state media-affiliated account says AI model developers DeepSeek and MiniMax have completed required registration steps and obtained permission to release and license their...
Chery unveils Ziwei in-house chip lineup for AI computing, SiC power and vehicle control
Chery unveils its Ziwei automotive chip lineup at its Global Innovation Conference on October 20. The company presents a portfolio spanning three categories: central computing chip...
Ant Group open-sources Ming-3.0 speech models spanning TTS, audio understanding and recognition
Ant Group has open-sourced its Ming-3.0 family of speech models, released on October 20. The release makes multiple models available for different tasks, including speech synthesis...
Alibaba DAMO Academy unveils XuanTie E908A automotive RISC-V processor IP
Alibaba’s DAMO Academy introduces the XuanTie E908A automotive processor IP at the RISC-V China Summit in Shenzhen on October 19. The IP is positioned for safety-related automotive...
SpacemiT Unveils V100 Server-Class RISC-V CPU With Validation on Multiple Linux Platforms
SpacemiT has announced its V100 server-class RISC-V processor and pairs it with three server systems. Reporting from the outlets describes the company as positioning V100 for serve...