pandaily

Saudi HUMAIN-M3 Builds on MiniMax M3 as Nations Hunt Open-Weight AI Bases

PIF-backed HUMAIN launched an Arabic frontier model on MiniMax's open MoE stack—showing how sovereign AI projects are turning Chinese open weights into national starting points. HUMAIN-M3 keeps M3's ...
pandaily

Global AI Compute Infrastructure Could Top $4 Trillion by 2028, Wuxi Summit Hears

At Wuxi's integrated-circuit innovation conference, industry executives projected global new data-center infrastructure investment will exceed $4 trillion by 2028, with semiconductor spending over ...
pandaily

GLM-5.3 Arrives Hours After Tang Jie's 'sooooooon' — Zhipu's Coding and Security Model Doubles SWE-Marathon, Tops CyberGym

Three days after a user teased Zhipu AI chief scientist Tang Jie on X about GLM-5.3, he replied 'sooooooon', and hours later the model was live. Focused solely on coding and security, GLM-5.3 more ...
pandaily

China's 'Thinking Machines': VUI Labs' Luna-TTS Tops the Global TTS Arena, Beating ElevenLabs and MiniMax

Chinese voice AI startup VUI Labs has taken its Luna-TTS model to number one on Hugging Face's TTS Arena, beating ElevenLabs, MiniMax, and Cartesia, and to third place on Artificial Analysis' Speech ...
pandaily

5G Private Networks at an Inflection Point: Omdia Says Niche Growth, AI and Sovereignty Are Keys, Challenges Remain

Omdia's 2026 end-to-end 5G private network vendor radar finds the market at a turning point: some operators held unrealistic expectations and failed to deliver, while innovation continues. The market ...
pandaily

A Local Opus? Alibaba Qwen Open-Sources Qwen3.8-27B — Frontier Coding and Agent Scores That Runs on 17GB of RAM

Want to read in a language you're more familiar with? Alibaba's Qwen team open-sourced Qwen3.8-27B, which topped Hugging Face's global trending chart within two days and passed one million downloads.
pandaily

The Hottest Embodied-AI Year Yet, Yet Robot Evolution Is Hard to See With the Naked Eye

At the 2026 World Robot Conference, visitors saw familiar scenes: robots grasping cups, folding clothes, entering factories. The real changes hide beyond the spectacle. Global humanoid shipments ...
pandaily

China's Dexterous Hand New Top 4: From Demo to Mass Production, the First Tier Solidifies

Want to read in a language you're more familiar with? Yinshi Robotics, Lingxin, LinJieDian and BrainCo form the new top tier as Chinas dexterous hand industry crosses the demo-to-mass-production ...
pandaily

MiniMax H3 Open-Sourced: H3-Omni-Transformer With 33B Parameters Benchmarks Against Seedance 2.0 at One-Third API Price, Runs Locally on Two RTX 5090 Cards

Want to read in a language you're more familiar with? MiniMax releases H3-Base weights on HuggingFace, with 33B dense single-stream Omni-Transformer, 16x spatial and 4x temporal compression VAE, and ...
pandaily

DeepSeek-V4-Flash Official Version Released and Open-Sourced: 304B Lightweight Model Surpasses V4-Pro Preview and Matches Claude Opus-4.8, Hailed as Third DeepSeek Moment

DeepSeek-V4-Flash-0731 official release outperforms V4-Pro preview with 82.7 performance score at 0.14/0.28 USD pricing, tops VulcanBench, and reaches HuggingFace trending second. DeepSeek released ...
pandaily

CXMT LPDDR6 Hits Key Breakthrough: Research Validation Nearing Completion as ChangXin Memory Technologies Closes In on Mass Production of 12,800 Mbps Mobile DRAM

Want to read in a language you're more familiar with? CXMT LPDDR6 development validation nearing completion, 12,800 Mbps design rate, 16Gb density, positioning for H2 2026 mass production against SK ...
pandaily

Deconstructing the Kimi K3 Technical Report: Moonshot AI Starts Setting Problems for DeepSeek as Two Scaling Axes Redefine the Frontier

Moonshot AI 47-page K3 report reveals KDA+Gated MLA hybrid attention, AttnRes cross-layer retrieval, Stable LatentMoE with 896 experts, and 51.2M RL sandboxes. Kimi K3 47-page technical report reveals ...