DailyVOL 2026.09.21112 items

Multiple models reach 1M context length, AI hardware and products accelerate deployment

Singapore·Monday, 21 September 2026·DAILY · 08:00 SGT

2 Sep 2026. Multiple large language models have reached 1 million token context lengths, including DeepSeek V4.1-Flash, Step 5 Preview, and StepFun's Step 5 preview. These advancements in model capacity are being supported by new AI hardware deployments such as Digital China's KunTai supercomputer system featuring the Kunpeng 950 and Ascend 950DT chips. Meanwhile, AI products are seeing rapid adoption across industries, with Tesla reporting a 40% lower collision rate for its FSD (Supervised) system in Australia and New Zealand compared to human driving. Additionally, the FBI has experienced a 605% increase in AI usage, leveraging the technology to process communications and prevent violent incidents. The combination of improved model capabilities, advanced hardware, and practical product implementations is driving AI integration across various sectors.

01

Models

4 items
02

Products

4 items
03

Industry

4 items
04

Papers

3 items
05

How-to

4 items
112Stories
53First-party
54New models
34Sources
AITOP · generated by the editorial system