Home
Tools
Models
App
Prompts
News
Topics
Submit for inclusion
English
English 日本語 한국어 Português español Deutsch Русский Français 繁體中文 简体中文
option
Home Close
Home
Tools
Category Ranking Generate Cover Generate Image Skill
Models
Large Language Model Multimodal Model
App
Prompts
Prompts Image Prompts
News
News Flash News
Topics
Submit for inclusion
Submit App Submit Tool
Create an account Sign In
Change Language English Change Language
Settings
English EN 日本語 JA 한국어 KO Português PT español ES Deutsch DE Русский RU Français FR 繁體中文 ZH-TW 简体中文 ZH-CN
Home
Flash News
Content
FredScott
FredScott
May 20, 2026

Alibaba Cloud unveiled a 128-card ultra-node server powered by its new AI chip Zhenwu M890 at its 2026 summit. The server integrates 128 chips via a self-developed interconnect chip achieving nanosecond-level latency. This architecture aims to tackle extreme computing demands of the Agentic era focusing on massive Agent inference and large model training.

0
0
Share Click to copy the link
Comments (0)
0/300
Submit
Author RichardJackson RichardJackson September 20, 2026
Step launches Step5Preview, a 600B sparse MoE model optimized for real-world Agentic tasks. Activating only 27B parameters, it supports 1M token context and multimodal inputs. Scoring top-three globally, it costs one-eighth of Claude Opus5 while matching performance in coding, long-horizon reasoning, and finance. Key benchmarks include surpassing Opus5 in GPU kernel optimization and demonstrating robust multi-step agent workflows. Full open-source weights release October 15, positioning Step5Preview as a high-efficiency flagship for daily agent use.
0
0
/live/7474
Author KeithNelson KeithNelson September 20, 2026
Thunder launches AI Master M7000, a 120B parameter local AI workstation priced at 19,900 RMB. Featuring AMD Ryzen AI Max+395 and 64GB RAM, it utilizes Phison's aiDaptiv+ to treat SSD as GPU memory extension. This enables efficient local inference of large models without professional GPUs, targeting creation, development, and confidential work in a 2.45kg device.
0
0
/live/7473
Author GregoryWalker GregoryWalker September 20, 2026
JumpStellar launches Step 5 Preview, a 60B sparse MoE model rivaling 2.8T parameter performance with high cost efficiency. It matches Kimi K3 Max on benchmarks, offers 1M token context, and autonomously optimizes GPU kernels for 24 hours. Available via API and Studio, new users get up to 75 days free. Performance approaches DeepSeek V4.1 Flash, though speed may vary with usage volume. Full-weight model releasing October 15.
0
0
/live/7470
Author MarkWilson MarkWilson September 20, 2026
GPT-6Astra and three researchers solved a 2017 math problem on FrontierMath proving the core always exists in approval-based committee elections. The AI invented a new voting rule using harmonic entropy and provided a polynomial-time algorithm. This marks the first time AI solved a major breakthrough math problem, shifting large models from problem-solving tools to collaborative pattern-discoverers. Epoch AI now recognizes Human+AI joint achievements, signaling a rewrite of scientific research methods where humans act as prompt engineers guiding AI-driven discovery.
0
0
Author DanielPerez DanielPerez September 20, 2026
Bilibili launched AI Infinite Arena, a real-world model competition. GPT-6Astra ranked first in initial tests by UPs, followed by GLM-5.3. Over 100 models including DeepSeek and ChatGPT compete across dynamic topics like coding and reasoning without fixed metrics. Rankings update in real time as creators submit challenges, highlighting strong performance by domestic models against global giants
0
0
/live/7468
OR