DeepSeek Slashes AI Model Prices by 97%, Igniting Market Competition

DeepSeek has significantly reduced the prices of its artificial intelligence (AI) models, including its latest V4 offering, which now costs 97% less than comparable OpenAI products. This strategic move aims to attract more enterprise clients, developers, and agent-based users.

Credit: UNSPLASH

The company announced permanent price cuts for "input cache hits" for application programming interface (API) users, reducing costs to one-tenth of original levels. The minimum input cost is now approximately USD 0.14 per million tokens, effective immediately.

To promote its new V4-Pro model, DeepSeek also offered an additional 75% discount through May 5. This makes DeepSeek-V4-Pro as inexpensive as USD 0.0036 per million input tokens, significantly lower than its American rivals.

In comparison, OpenAI’s GPT-5.5 charges USD 0.5 per million cached input tokens. A conversation on GPT-5.5 can be 32 times more expensive than on DeepSeek-V4, considering input typically exceeds output in length.

DeepSeek's aggressive pricing reflects heightened competition within China's foundational model market. Other high-profile start-ups, such as Kimi K2.6 and Zhipu GLM-5.1, recently increased prices on their latest flagship versions.

DeepSeek V4 is one of the few models to break this trend, prompting speculation of further price competition. Following DeepSeek-V4's release, OpenRouter, a US-based model aggregation platform, reported a notable surge in usage.

DeepSeek-V4-Pro recorded 13.6 billion tokens recently, nearly four times its previous day’s volume on OpenRouter. Hu Yanping, a distinguished professor at Shanghai University of Finance and Economics, noted DeepSeek's goal of attracting more users and lowering industry pricing expectations.

Professor Hu also suggested that DeepSeek's price cuts might have a relatively limited impact on top-tier models like GPT-5.5 and Claude 4.7 Opus. DeepSeek stated its V4 models are optimised for mainstream agent tools.

Users must configure Claude Code by setting the model to "deepseek-V4-pro" for 1-million-token context support, while OpenCode and OpenClaw require an upgrade. This pricing overhaul accompanies technical upgrades and deeper integration with Huawei Technologies’ Ascend ecosystem.

This integration is expected to further enhance cost efficiency. DeepSeek-V4 is viewed as enabling more complex agent-based applications at a lower cost, expanding the scope for scalable AI deployment.

A report from Goldman Sachs highlighted that integration with Ascend supernodes could strengthen DeepSeek's cost competitiveness and support wider adoption. Arena.ai evaluated DeepSeek-V4-Pro's performance as comparable to GPT-5.4-high and Gemini-3.1-Pro in agent-based web development tasks.

Artificial Analysis found DeepSeek-V4 to be far more cost-efficient than leading closed-source models such as Claude, despite performance lags. On the third-party benchmark firm’s flagship Intelligence Index tests, DeepSeek’s Flash variant cost approximately USD 113 in inference terms.

In contrast, Anthropic’s latest Claude Opus 4.7 model incurred costs around USD 4,811 for the same tests.

---

• DeepSeek has reduced prices for its V4 AI models by 97% compared to OpenAI's GPT-5.5.

• The V4-Pro model is now available for as low as USD 0.0036 per million input tokens, with a 75% promotional discount through May 5.

• DeepSeek's pricing strategy aims to attract enterprise clients, developers, and agent-based users amidst increasing competition in China's AI market.
  

Source: SCMP

OpenAI AI Smartphone Processor Collaboration Fuels Qualcomm Share Surge

Qualcomm shares rose 13% in premarket trading following an analyst report about OpenAI's collaboration with the chip designer and Taiwan's MediaTek. The partnership aims to develop smartphone processors.

Credit: Ariana Drehsler/Bloomberg

TF International Securities analyst Ming-Chi Kuo reported that OpenAI, the creator of ChatGPT, plans an AI-first smartphone. Qualcomm and MediaTek are listed as co-development partners for this device.

Mass production for the AI-first smartphone is anticipated in 2028. Kuo shared this information on social media platform X.

China's Luxshare, an Apple supplier, is the exclusive system design and manufacturing partner for the device, according to Kuo. The analyst, based in Taiwan, is known for his accurate predictions regarding Apple products.

OpenAI has explored consumer AI devices for years. The organisation last May acquired Jony Ive's startup io Products for $6.5 billion. Ive, a former Apple designer, was tapped to lead these efforts.

However, media reports have suggested the proposed device may not be a smartphone. Altman reportedly informed employees that the device would be a "third core device," alongside phones and laptops, according to The Wall Street Journal.

The startup, which is currently loss-making, has also reduced its focus on side projects. Instead, it is concentrating on coding tools for businesses, an AI area demonstrating clear commercial traction.

Launching a smartphone would position OpenAI directly against deep-pocketed rivals such as Apple and Samsung. These two companies collectively hold approximately a 40% share of the global smartphone market.

Such a move would also add to signs that the smartphone would likely retain its central role in daily life during the AI era. Reuters reported last month that Amazon is also planning a renewed push into the handset market.

Apple shares decreased by 1.7%. The company last week named long-time hardware chief John Ternus as CEO. This is a sign that devices would continue to play a central role in Apple's business, even as it looks to catch up in offering AI to users.

---

• Qualcomm shares surged 13% in premarket trading on Monday after a report detailing an OpenAI partnership with the chip designer and Taiwan's MediaTek for AI smartphone processors.

• OpenAI is reportedly collaborating with Qualcomm and MediaTek to develop processors for an AI-first smartphone.

• Mass production for the device, for which Luxshare is the exclusive system design and manufacturing partner, is likely in 2028.

Source: REUTERS

China Orders Meta to Unwind USD 2 Billion AI Startup Acquisition

China ordered U.S. tech giant Meta to unwind its acquisition of artificial intelligence startup Manus on Monday, a deal valued at more than USD 2 billion. This directive comes as Beijing intensifies scrutiny of foreign investment in domestic startups developing frontier technologies.

Credit: UNSPLASH

The National Development and Reform Commission (NDRC) stated its commitment to preventing U.S. firms from acquiring Chinese AI talent and intellectual property. This action aligns with Washington’s efforts to restrict Chinese tech firms’ access to advanced U.S. chips.

The NDRC’s office for reviewing foreign investment security announced it would "prohibit foreign investment in Manus in accordance with laws and regulations, and requires the parties involved to withdraw the acquisition transaction." Meta was not explicitly named in this statement.

After a USD 75 million fundraising round led by U.S. venture firm Benchmark in May 2025, Manus shut its China offices in July, laying off dozens of employees. It then moved its operations to Singapore.

This move allowed Manus’s parent company, Butterfly Effect, to re-incorporate in Singapore. The restructuring aimed to bypass U.S. investment restrictions on Chinese AI firms and Chinese rules limiting domestic AI firms’ ability to transfer intellectual property and capital overseas.

California-based Meta stated the transaction fully complied with applicable law and expressed anticipation of an appropriate resolution to the inquiry, in response to Beijing's latest move on Monday. The company acquired Manus to enhance its work on AI agents, which are tools designed to carry out complex tasks with minimal human intervention.

China’s commerce ministry announced a probe into the sale in January, days after Meta completed its acquisition. A ministry spokesperson previously stated that companies involved in foreign investment, technology exports, data transfers abroad, and acquisitions must comply with Chinese laws and regulations.

Xiao Hong, Manus’s chief executive officer, and Ji Yichao, its chief scientist, were summoned to Beijing for talks with regulators in March. Both co-founders were later barred from leaving the country, according to five sources familiar with the matter.

Despite the exit bans on its executives, Manus staff have relocated to Meta’s Singapore offices, and projects are proceeding, according to two sources familiar with the matter. Analysts and lawyers said the rare move to unwind a completed deal underscores Beijing’s intent to assert jurisdiction over cross-border transactions involving Chinese assets, shareholders, or technology under its national security review regime.

Weiheng Chen, senior partner and head of Greater China at law firm Wilson Sonsini, noted that China’s national security clearance will become "a regular closing condition for cross-border tech deals."

Carl Li, a partner with Chinese law firm Zhong Lun, commented in a post on his LinkedIn page on Monday that regulatory analysis of an acquisition is no longer confined to the target company’s place of incorporation. He added that the origin of technology, location of core research and development, nationality and location of the founding team, historical China operations, data flows, and offshore restructuring process may all become relevant.

Li further suggested that in sensitive technology sectors, a deal may be reviewed not only as a merger and acquisition transaction. It could also be considered a potential transfer of strategic technology, data, know-how, and national security-sensitive capabilities.

This order is China’s latest high-profile action to challenge a cross-border transaction involving a non-China incorporated company, amidst geopolitical tensions with Washington. China previously criticised Hong Kong billionaire Li Ka-shing’s CK Hutchison for agreeing a USD 23 billion sale of ports worldwide to a consortium led by U.S. asset manager BlackRock.

The NDRC’s decision serves as a significant warning to Chinese startups, especially in sensitive sectors like technology, that seek to move operations to Singapore to access foreign capital. This practice is often referred to as "Singapore washing."

Ben Chester Cheong, a lecturer at the Singapore University of Social Sciences, stated this action raises the compliance threshold rather than ending such moves. He suggested companies may need to demonstrate a genuine operational shift, including where management sits, intellectual property is owned, research and development is conducted, data is stored, and whether Chinese regulatory approvals are needed.

Manus was hailed early last year by state media and commentators as China’s next DeepSeek after releasing what it said was the world's first general AI agent. The company develops an agent framework that operates on top of existing Western large language models, rather than building its own AI model.

Alfredo Montufar-Helu, a managing director at Ankura China Advisors, highlighted that AI has become central to strategic competition between the world’s two largest economies. He explained that China is asserting it will prevent foreign acquisition of assets deemed important for national security, with AI now clearly among them.

---

• China has ordered U.S. tech giant Meta to unwind its acquisition of AI startup Manus, a deal exceeding USD 2 billion.

• The National Development and Reform Commission cited concerns over foreign acquisition of Chinese AI talent and intellectual property.

• Manus had relocated its operations and parent company to Singapore, attempting to bypass investment restrictions.

Source: REUTERS