Search results for "FLASH"
22:12

Microsoft Open Source New Version of Phi-4: Inference Efficiency Rises 10 Times, Can Run on Laptops

Jin10 data reported on July 10, this morning, Microsoft open-sourced the latest version of the Phi-4 family, Phi-4-mini-flash-reasoning, on their official website. The mini-flash version continues the Phi-4 family's characteristic of small parameters and strong performance, specifically designed for scenarios limited by Computing Power, memory, and latency. It can run on a single GPU, making it suitable for edge devices such as laptops and tablets. Compared to the previous version, mini-flash utilizes Microsoft's self-developed innovative architecture SambaY, achieving a 10-fold increase in reasoning efficiency, with average latency reduced by 2 to 3 times, resulting in a significant overall improvement in reasoning performance.
More
08:26

Frequent wildfires: Canada attempts a new method of seed-planting reforestation.

Canada is using drones to disperse seed capsules in an effort to accelerate vegetation recovery after wildfires. These capsules contain tree seeds and nutrients, and are implemented in collaboration with the local community by the company "Flash Forest". This method is more efficient than traditional tree planting.
More
01:29

The demand in the enterprise market has slowed in Q1, with the global NAND market size decreasing by 25% quarter-on-quarter.

Jin10 data reported on May 21, after several consecutive quarters of high rise, the procurement demand for enterprise SSDs in the first quarter showed a month-on-month decrease due to the delay of some data center projects. The stalling of enterprise SSDs as a key growth engine has broken the pattern of "server demand hedging against weak consumer end," resulting in a very weak overall demand for NAND Flash in Q1, with a clear oversupply in the market. In the first quarter, the global NAND Flash market size reached $13.01 billion, decreasing by 25.3% month-on-month and 14.2% year-on-year.
More
  • 1
04:21
On June 5th, Jinshi Data reported that this morning, Zhipu AI released version 2.0 of the MaaS platform open platform, including a series of new features and models. According to the introduction, the new GLM-4-Flash model masters 26 languages, with a comprehensive ability increase of 40%, surpassing the Llama-3-8B-Instruct in all aspects.
GLM-2.65%
03:59
On June 5th, Jins data reported that Zhipu AI announced a decrease in the prices of all model matrices, with GLM-4-Air and GLM-3-Turbo priced as low as 0.6 yuan per million Tokens, and the Embedding-2 model as low as 0.3 yuan per million Tokens. The GLM-4-Flash model has the highest price drop to 0.06 yuan per million Tokens.
GLM-2.65%
TURBO-5.28%
  • 2
06:32

TrendForce: The revenue of the top five NAND Flash brands is expected to reach 12.02 billion USD in the first quarter of 2025.

On May 29, Jin10 reported that according to the latest research by TrendForce, in the first quarter of 2025, NAND Flash suppliers are facing inventory pressure and declining demand from end customers, resulting in a 15% quarter-on-quarter decrease in average selling price (ASP) and a 7% decrease in shipment volume. Even though some product prices rebounded at the end of the quarter, boosting demand, the total revenue of the top five NAND Flash brands amounted to 12.02 billion USD, a nearly 24% quarter-on-quarter decrease.
More
00:29

Kai Xia's production capacity is fully loaded, and it is reported that the most advanced NAND Flash products will be mass-produced in July.

On July 4th, Jinshi Data reported that the operating rate of the Armor Hero production line has returned to 100% level in June, and it will mass produce the most advanced storage chip (NAND Flash) products in July to meet the increasing demand for data storage due to the popularity of generative AI. It is reported that Kioxia will begin mass production of NAND.
More
02:14

DeepSeek launches Open Source week

Jinshi data news on February 24th, DeepSeek launched the 'Open Source Week' today, the first Open Source code repository is Flash MLA - an efficient MLA decoding core optimized for Hopper GPU, designed specifically for processing variable length sequences. According to reports, the inspiration for Flash MLA comes from the FlashAttention 2&3 and cutlass projects.
More
DEEPSEEK-2.77%
  • 2
  • 1
04:14

Institution: The expected capacity of AI SSD purchases is projected to exceed 45EB in 2024.

AI service demand is rising, driving the demand for enterprise-level solid-state drives (Enterprise SSDs) to rise. Upstream suppliers are planning to launch 2YY products to meet the market demand for AI applications, and mass production is expected to start in 2025. It is estimated that the purchasing capacity of AI-related SSDs will exceed 45EB this year, and in the coming years, AI servers are expected to drive an average annual growth rate of SSD demand of over 60%. At the same time, the proportion of AI SSD demand in the entire NAND Flash is expected to rise from 5% in 2024 to 9% in 2025.
More
04:51

Institutions: It is expected that NAND Flash prices will drop by over 10% in the first quarter of next year.

On December 31, Jinshi Data learned that, according to the latest survey by TrendForce, NAND Flash suppliers in the first quarter of 2025 will face challenges such as continuous rise in inventory and declining order demand, with an average contract price expected to decrease by 10% to 15% per quarter. Among them, the decline in wafer prices will converge. For some module products, due to the stable orders of Enterprise SSD, it is expected to buffer the decline in contract prices; Client
More
05:46

Institution: NAND Flash contract prices are expected to drop by 3% to 8% in Q4.

A survey shows that the contract price of NAND Flash products will decrease by 3% to 8% in the fourth quarter of 2024, with the wafer contract price expected to drop by more than 10%. As for module products, Enterprise SSD may rise by 0% to 5%, but PC SSD and UFS will be affected by lower-than-expected sales of buyer terminal products, and the outlook for PA is even more conservative.
More
12:25

GigaDevice: The impact of general NOR Flash products on the company's revenue this year is relatively small

On December 20, Jinshi Data reported that Meguiar Innovation disclosed in the evening announcement of stock trading anomalies that recently, the news about the application of NOR Flash products in the AI-related terminal market has attracted high attention from investors. The company explained the following about the relevant situation: The company's general NOR Flash products can be widely used in consumer electronics, intelligent wearable devices and other terminal fields, including AI headphones, AI glasses, AI PCs and other fields. However, due to the company's provision of general NOR
More
  • 1
12:48

Puran Shares: The direct impact of AI headset-related business on the company is uncertain.

On December 13th, Jinshi Data noticed that the 'AI headphone concept' has received high attention in the market. The company would like to explain the following related information: General NOR Flash products can be widely used in wearable markets such as headphones, watches, and bracelets, as well as toys, smart AIOT, and other terminal products. The company serves as a general NOR
More
  • 2
  • 2
04:40

Institutions: Estimated third-quarter NAND Flash contract price increase narrowed to 5-10%.

TrendForce's latest survey shows that the oversupply ratio of NAND Flash in the third quarter has risen to 2.3%, and the average price increase of NAND Flash has converged to a quarterly increase of 5-10%. Enterprises continue to invest in server construction, especially Enterprise SSDs benefiting from the expanded use of AI, continue to be driven by orders, while demand for consumer electronics remains weak.
More
  • 1
Trade Crypto Anywhere Anytime
qrCode
Scan to download Gate app
Community
English
  • 简体中文
  • English
  • Tiếng Việt
  • 繁體中文
  • Español
  • Русский
  • Français (Afrique)
  • Português (Portugal)
  • Bahasa Indonesia
  • 日本語
  • بالعربية
  • Українська
  • Português (Brasil)