DeepSeek V4 Launches with Enhanced Reasoning Capabilities

Artificial Intelligence

The Fourth Industrial Revolution

Sam Mendoza

·2 years ago

shared a link post in group #Artificial Intelligence

DeepSeek V3, the latest release from #🇨🇳 ChinA.I. 🤖🧠🦾🤖  firm DeepSeek, has set a new standard for open #Artificial Intelligence  models. With an impressive 671 billion parameters, it outperforms competitors like Meta’s Llama 3.1 and OpenAI’s GPT-4o in tasks such as coding, translation, and writing. What’s even more remarkable is its training cost—just $5.5 million, a fraction of what similar models require. However, DeepSeek V3’s design reflects Chinese regulatory influences, avoiding discussions on politically sensitive topics to align with “core socialist values.”

https://techcrunch.com/20..

DeepSeek's new AI model appears to be one of the best 'open' challengers yet | TechCrunch

techcrunch.com

DeepSeek's new AI model appears to be one of the best 'open' challengers yet | TechCrunch

ChCC

Jerome Cody
Hangzhou-based firm is recruiting developers to test software that turns language models into autonomous agents, escalating price war with US rivals #🇨🇳 ChinA.I. 🤖🧠🦾🤖 company DeepSeek is inviting open-source developers to test its upcoming “harness” – software designed to turn large language models (LLMs) into #Artificial Intelligence agents – accelerating a push into agentic tech as DeepSeek’s latest V4 Flash model sends another cost-efficiency shock wave through Silicon Valley. The Hangzhou-based firm was looking for open-source project developers to join beta testing for DeepSeek Harness, according to a social media post on Saturday by Cui Tianyi, who leads the harness team. #(Cool) Dev Agentic harness frameworks – designed to orchestrate LLMs so they can execute multi-step code, reason through complex workflows, and act autonomously – have become a major battleground for artificial intelligence labs following the commercial success of Anthropic’s Claude Code. DeepSeek hired Cui in March to join its newly formed harness group. He was previously a Hong Kong-based co-founder of quantitative trading firm TSY Capital and a software engineer at quantitative trading firm Jane Street. Cui did not disclose a release date for DeepSeek Harness. However, he noted on social media in June that while the team held ambitious goals, it remained severely short-staffed and was recruiting across multiple channels. The agent push comes as CEO and co-founder Liang Wenfeng doubles down on cheap, highly capable models as a core strategy towards artificial general intelligence (AGI), a coveted milestone where AI systems match or surpass human cognitive abilities. scmp.com/tech/tech-trends..
www.scmp.com
China’s DeepSeek tests AI ‘harness’ as cheap V4 model jolts Silicon Valley
Hangzhou-based firm is recruiting developers to test software that turns language models into autonomous agents, escalating price war with US rivals.

Bobby
Jerome Cody
OpenAI and Microsoft may have blocked th...
Accusations against Chinese tech companies of illicit activity are not new, especially from the US. Still, the last few days have seen a remarkable escalation in how Washington and Silicon Valley are calling out China’s AI industry, most notably on the controversial issue of AI model distillation — the alleged practice of Chinese companies “stealing” the outputs of US AI models to train their own. On Tuesday, three US government agencies including the FBI and the NSA accused six Chinese companies of engaging in an industrial campaign to steal US model outputs, “likely with Chinese government awareness”. The report from leading US law enforcement and intelligence agencies was soon followed by a report from leading US AI company Anthropic, which claimed that its Chinese AI rivals including national champions DeepSeek and Moonshot covertly routed customer requests to its Claude models and passed those responses as their own. #🇨🇳 ChinA.I. 🤖🧠🦾🤖 AI and legal experts say that the claims mark a new intensity in US criticism of China’s AI industry, given the unprecedented detail provided in the reports and the significance of the agencies behind them. What is notable is that Chinese AI companies accused of such malpractice have largely stayed mum about these allegations. Instead, they have left it to officials in Beijing to respond, with China’s Ministry of Commerce saying on Wednesday that Washington is using these allegations to unfairly suppress China’s AI industry. scmp.com/economy/article/..
www.scmp.com
China rejects US AI model distillation allegations, warns of retaliation
The commerce ministry says distillation is ‘neutral’ and widely used by companies around the world, including American firms.

Cyril
DeepSeek is raising at least $12 billion in Tencent-backed funding. This would provide the Chinese AI lab with a strong war chest to challenge its better-capitalized rivals in the West. — Bloomberg (substack.com/redirect/b4a..) #🇨🇳 ChinA.I. 🤖🧠🦾🤖
substack.com
DeepSeek to Raise at Least $12 Billion in Tencent-Backed Funding
DeepSeek is close to securing at least 80 billion yuan ($12 billion) in its latest round of funding, blowing past its own capital-raising target ahead of a landmark initial public offering in early 20

Jozhe took a screenshot of this post and discussion

Cyril
DeepSeek has open-sourced DeepInfer, the in-house framework for running its own models. The move reveals some of the technical secrets behind how the Chinese AI developer achieves high efficiency and low compute costs when running its models at scale. #🇨🇳 ChinA.I. 🤖🧠🦾🤖 Access to DeepInfer is likely to help third-party AI inference platforms that host DeepSeek models, such as Fireworks and Baseten, run DeepSeek’s latest models more efficiently and earn higher margins. But DeepInfer could also have a broader impact. Other AI labs in China and the U.S. could study DeepSeek’s open-source framework and learn how to increase the inference efficiency of their own models, developers say. DeepInfer supports Nvidia’s Blackwell chips as well as Huawei’s Ascend 950 chips, DeepSeek said. DeepInfer, however, doesn’t reveal everything. DeepSeek, for example, hasn’t open-sourced some of its proprietary techniques, such as those for managing memory and controling how the model distributes traffic across servers. While DeepSeek’s revenue is minuscule compared to OpenAI and Anthropic, the Chinese company is achieving healthy margins. Its gross margin for the business of selling access to its models in the first seven months of this year was 82.9%, higher than both Anthropic and OpenAI during similar periods github.com/deepseek-ai/De..
github.com
GitHub - deepseek-ai/DeepInfer: DeepInfer: High-performance, large-scale inference framework for DeepSeek models
DeepInfer: High-performance, large-scale inference framework for DeepSeek models - deepseek-ai/DeepInfer

Embed post to a webpage :

<div data-postid="bxawgae" [...] </div>