Breaking news

DeepSeek Releases V4 Model With Lower Inference Costs

Chinese AI startup DeepSeek has introduced a preview version of its highly anticipated V4 large language model. The new release, following the success of its R1 reasoning model last year, underscores DeepSeek’s intent to disrupt the global AI landscape by delivering robust performance at substantially reduced computational costs.

Preview Of V4 LLM And Its Enhanced Features

The V4 model is available in “pro” and “flash” versions, designed for different deployment scales. It is open-source, allowing developers to access, run and modify the model. Neil Shah, Vice President of Research at Counterpoint Research, said the preview demonstrates strong capabilities in agent-based tasks, knowledge processing and inference efficiency. Lower inference cost, the compute required to generate outputs, remains a key factor in reducing deployment expenses.

Competitive Dynamics And Market Response

DeepSeek’s earlier R1 model intensified competition with global and domestic players, including Nvidia and Google, by highlighting cost-efficient model scaling. While V4 is not expected to deliver the same level of disruption, its release reinforces competition within China’s AI sector. Companies such as Alibaba and ByteDance continue to introduce new models, indicating sustained investment and rapid iteration across the market.

Chip Technology And Implications For AI Sovereignty

Huawei said its AI compute cluster, powered by Ascend processors, supports V4. The extent to which domestic chips are replacing hardware from Nvidia remains unclear, as US export restrictions continue to limit access to advanced semiconductors. Analysts at Counterpoint Research said optimization for Chinese chipmakers, including SMIC and Hua Hong Semiconductor, could support Beijing’s push for greater technological independence.

Market Outlook

DeepSeek’s V4 release highlights ongoing shifts in AI development, with cost efficiency and domestic infrastructure becoming central to competition. Increasing activity among Chinese developers continues to reshape both local and global market dynamics.

Eurobank Plans €1 Billion Investment In AI And Digital Banking By 2028

Eurobank plans to invest about €1 billion in technology from 2025 through 2028, its largest technology investment program to date. The Banking Forward strategy focuses on digital banking, artificial intelligence, customer experience and a “phygital” model combining digital services with face-to-face support.

Digital Banking Dominates Customer Activity

Digital channels already account for 96% of Eurobank transactions, with 61% completed through the Eurobank Mobile App. Among customers aged 35 and under, digital adoption reaches 94%.

Customers make about 574 million annual logins across e/m-banking and more than 1 million digital transactions each day. During the first half of 2026, one in three banking products was acquired digitally.

AI Moves Into Everyday Banking

Eurobank is expanding the use of AI through tools including EVA, its digital customer assistant, and myEVA, an AI-powered voice assistant for employees. The technology is also being applied to mortgage assessments, customer feedback analysis and contractual documents.

The bank’s technology architecture is built around five areas: digital channels, customer experience orchestration, data and AI, core banking, and infrastructure and cloud. About 50% of its applications and digital channels are already cloud-based.

Investment Extends Beyond Technology

The program is intended to reshape how Eurobank operates, combining automation and AI with employee development and human support. The bank says the approach is designed to improve services while maintaining access to face-to-face banking when customers need it.

Aretilaw firm
The Future Forbes Realty Global Properties
Uol
eCredo

Become a Speaker

Become a Speaker

Become a Partner

Subscribe for our weekly newsletter