appConfigEntityId

bodyType

customContents

entityId

entityUuid

isVideo

articlePageQuery

article

articleSchemaContent

content

hooksArticlePageAdvertisementConfigArticle

hooksArticlePageFloatingButtonConfigArticle

appConfig

filter

json

liveArticle

Xinmei Shen

Chinese artificial intelligence start-up DeepSeek has quietly released DeepSeek-V4-Pro-0813, an updated version of its latest flagship model, leaving some developers underwhelmed by its overall capabilities and disappointed in its pricing – but impressing researchers in niche areas like cybersecurity.
The stealth update to April’s preview version came with a brief statement on DeepSeek’s official website on Wednesday, noting that the model offered “significantly enhanced agent capabilities”, but the statement was removed as of Thursday afternoon. The release follows the company’s smaller V4 Flash model, which jolted Silicon Valley last month due to its extreme cost efficiency.
However, early benchmark results suggest the new flagship is struggling to match its top-tier rivals.
DeepSeek-V4-Pro-0813 scored 53 on the Artificial Analysis Intelligence Index, on par with Zhipu AI’s GLM-5.2 released in June, but was four points behind the mid-tier Terra model in OpenAI’s latest GPT-5.6 series and seven points behind Moonshot AI’s Kimi K3.
On the Vals Index, compiled by San Francisco-based Vals AI to evaluate models across several benchmarks, the new DeepSeek model ranked 12th. It trailed OpenAI’s previous-generation GPT-5.5 and also lagged well behind frontier systems like Kimi K3 and Anthropic’s Claude Opus 5.
The model struggled in two areas in particular: completing tasks within a sandboxed terminal environment and generating complex financial models in Excel spreadsheets, Vals AI said on Wednesday.
On social media, the early findings prompted some developers and industry observers to describe the Hangzhou-based company’s model as “disappointing,” with several users noting issues such as the model stopping prematurely during long coding tasks.
DeepSeek did not immediately respond to a request for comment on Thursday.
Despite its mixed reception, the model’s cybersecurity performance stood out. According to Belgian cybersecurity firm Aikido Security, DeepSeek-V4-Pro-0813 outperformed every other model tested at finding system vulnerabilities.


While the model suffered from “poor precision”, it detected far more vulnerabilities than Anthropic’s Opus 5 and Alibaba Group Holding’s Qwen 3.8, Aikido researcher Philippe Dourassov said on Thursday.
Alibaba is the owner of the South China Morning Post.
Unlike typical AI lab releases, DeepSeek has yet to publish a technical blog post for the new model.
Pricing for the official Pro model has also drawn mixed reactions, particularly following the praise lavished on the ultra-cheap V4 Flash model released on July 31.
DeepSeek-V4-Pro-0813 is priced at about 44 US cents per million input tokens, which Artificial Analysis characterised as “somewhat expensive” compared with the market median of 33 US cents – and 87 US cents per million output tokens, described as “moderately priced.”
At 6 US cents per task, according to the Artificial Analysis Intelligence Index, the model is slightly more expensive to run than OpenAI’s lightweight GPT-5.6 Luna at 5 US cents, though it remains nearly 93 per cent cheaper than Moonshot AI’s Kimi K3 at 84 US cents.
DeepSeek last week said it was implementing a “significant” price increase due to surging demand, advising users to plan their usage accordingly.
Meanwhile, DeepSeek is expected to soon launch a “harness” product – a software framework similar to Claude Code that enables large language models to function as autonomous AI agents.
Earlier this month, the company invited open-source developers to join beta testing for the coming DeepSeek Harness, after it registered a WeChat account for the team in July.


Philippine tribe vows to stand ground against US-led Pax Silica: ‘listen to us’

As the US weighs restrictions on cloud computing, how will China’s AI sector adapt?

Chinese AI chipmaker Hygon plots expansion from data centres to robotics

For China’s hi-tech IPOs, it’s Hong Kong vs the mainland – but how much does it matter?

Artificial intelligence (AI)

DeepSeek’s updated V4 Pro AI model struggles on benchmarks, shines in cybersecurity

Chinese artificial intelligence start-up DeepSeek quietly released an updated version of its latest flagship model, DeepSeek-V4-Pro-0813, on Wednesday. Photo: Reuters

Meta to challenge China’s open-weight AI dominance amid US regulatory fears

Meta CEO Mark Zuckerberg attends the annual Allen and Company Sun Valley Media and Technology Conference at the Sun Valley Resort in Sun Valley, Idaho on July 9, 2026. Photo: Reuters

Meta Platforms has renewed its push into open-source models in a bid to lure away American customers who are worried about looming regulatory restrictions in Washington due to the adoption of Chinese AI. Photo: Reuters

China’s top AI is still trained on Nvidia chips. What is delaying a switch to local tech?

Huawei Technologies’ Compute Architecture for Neural Networks requires developers to rewrite and optimise large amounts of code. Photo: Shutterstock

Visitors learn about Huawei’s Ascend AI-based CloudMatrix 384 at the 2025 World AI Conference in Shanghai. Photo: Xinhua

OpenAI blinks in face-off with Chinese rivals, drops pricing for some models up to 80%

OpenAI CEO Sam Altman arrives for a meeting at the White House in Washington on July 30. Photo: Reuters

People visit the Moonshot AI booth during the World Artificial Intelligence Conference in Shanghai on July 18. Photo: AFP

Alibaba tests paid AI appetite with US$30 annual office-assistant subscription

The e-commerce giant’s Qwen app now offers three membership tiers for the office assistant. Photo: Shutterstock

Tencent ramps up overseas AI push with wider roll-out of flagship Hy3 model

The Tencent logo. Photo: Reuters

The Tencent booth during the World Artificial Intelligence Conference in Shanghai on July 18. Photo: AFP

Tech

Big Tech

Tech Trends

Tech War

Chinese AI start-up’s model impresses in niche areas like cybersecurity but leaves some developers disappointed in its overall capabilities


Chinese AI start-up’s model impresses in niche areas like cybersecurity but leaves some developers disappointed in its overall capabilities.


DeepSeek’s updated V4 Pro AI model struggles on benchmarks, shines in cybersecurity

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Chinese AI start-up’s model impresses in niche areas like cybersecurity but leaves some developers disappointed in its overall capabilities

Chinese AI start-up’s model impresses in niche areas like cybersecurity but leaves some developers disappointed in its overall capabilities