China's DeepSeek ups AI stakes as R2 rumours fly fast
Amid tariff turfs, China is rumoured to advance its supremacy in artificial intelligence (AI) model race by releasing
DeepSeek R2
, the second version of reasoning model which shook US stock markets in February.
#Pahalgam Terrorist Attack
India stares at a 'water bomb' threat as it freezes Indus Treaty
India readies short, mid & long-term Indus River plans
Shehbaz Sharif calls India's stand "worn-out narrative"
According to some reports in the Chinese media, DeepSeek R2, expected to be released this week, will be 97.3% cheaper than OpenAI's GPT-4o model and 100% trained on Huawei's Ascend 910B GPU cluster, establishing China's independence from American AI chips. Here's a look at what it will mean for the AI ecosystem
What will be the impact?
If the model performs on a par with competitors on global benchmarks, DeepSeek could establish Huawei as the first serious competitor to NVIDIA, market watchers wrote on X. The previous model released by the Hangzhou, China-based AI startup had wiped off $1.5 trillion off US stock markets, tumbling stocks of tech giants like NVIDIA, Alphabet and Microsoft.
Meanwhile, Meta, which has been an open-source category leader until now is also expected to lose market share with DeepSeek also offering its models free-to-access. This comes at a time when OpenAI is also looking to release its first open-source model soon.
Live Events
How big is the model?
In terms of size, DeepSeek R2 will be comparable to OpenAI's largest model so far GPT 4.5 code-named Orion with 1.8 trillion parameters. Meanwhile, DeepSeek R2 is expected to have 1.2 trillion parameters trained on 5.2 petabytes of training data.
Discover the stories of your interest
Blockchain
5 Stories
Cyber-safety
7 Stories
Fintech
9 Stories
E-comm
9 Stories
ML
8 Stories
Edtech
6 Stories
The pricing for consuming the model through APIs is expected to be $0.7 per million input tokens and $0.27 per million output tokens, which is 97.3% cheaper than OpenAI.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
24 minutes ago
- Time of India
No AI allowed: China shuts down DeepSeek and other AI chatbots for university entrance tests
Major Chinese technology companies have temporarily disabled key artificial intelligence features to prevent cheating during the country's high-stakes university entrance examinations. More than 13.3 million students are currently sitting the four-day "gaokao" tests, which began Saturday and determine their chances of securing limited university spots. Alibaba , Tencent , and ByteDance have suspended photo-recognition capabilities in their popular AI chatbots during exam hours. When students attempted to upload exam questions to apps like Doubao (owned by TikTok 's parent company) and Qwen, the services responded that question-answering functions were unavailable "to ensure fairness in the college entrance examination," according to Bloomberg reports. DeepSeek , a generative AI tool that gained prominence earlier this year, also blocked its services during specific hours for the same reason. Screenshots shared on Chinese social media platform Weibo showed frustrated students encountering these restrictions even when claiming their queries weren't exam-related. The AI suspensions represent just one layer of anti-cheating measures deployed during gaokao week. Several provinces have implemented AI monitoring systems to detect "abnormal behaviors" like whispering or suspicious glances between students during exams, The Guardian reported. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Medical Mystery Solved: Dementia and Memory Loss Has Been Linked To This Common Thing. vitalgethealth Click Here Undo Chinese authorities have also introduced stricter entry protocols, including biometric identification, enhanced screening for digital devices, and radio signal blockers at exam sites. The gaokao's significance extends beyond academics, some cities have postponed public events, delayed office hours, and created priority traffic lanes to ensure students arrive on time. For many Chinese families, particularly those from smaller cities and lower-income backgrounds, this exam represents the primary pathway to higher education and social mobility. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
India Gazette
37 minutes ago
- India Gazette
Indian Army to get Rs 30,000 crore QRSAM air defence missile system boost
ANI 09 Jun 2025, 20:47 GMT+10 By Ajit Dubey New Delhi [India], June 9 (ANI): At a time when Indian air defence systems successfully thwarted the Pakistani aircraft, missile and drones attacks under Operation Sindoor, the Indian Army is all set to get Rs 30,000 crore boost with a new surface-to-air missile system expected to be cleared for acquisition by the Defence Ministry soon. The Defence Ministry is scheduled to take up the proposal for buying three regiments of the indigenous Quick Reaction Surface to Air Missile system (WRSAM) for the Army Air Defence for deployment along both western and northern borders, defence officials told ANI here The missile systems, developed by the Defence Research and Development Organisation, are highly mobile as they have the capability to search and track targets on the move and fire at short halts, they said. With a range of around 30 km, the system would be complementing the existing systems in the forces like the MRSAM and Akash in short to medium ranges. The missile system performance has been extensively evaluated under both day and night operational scenarios during the trials. The meeting of Defence Acquisition Council is planned to be held around the fourth week of June. During the four-day conflict with Pakistan, which used Chinese weaponry, the Indian Army's Air Defence units destroyed majority of the drones using L-70 and Zu-23 air defence guns while the Akash and MRSAM played a crucial role along with the Indian Air Force's Spydwr and Sudarshan S-400 air defence systems. The Army Air Defence is also getting a number of new radars, very short range air defence systems along with jammers and laser-based systems to deal with drones of Turkish and Chinese origin. (ANI)
Time of India
an hour ago
- Time of India
Apple researchers find ‘major' flaws in AI reasoning models ahead of WWDC 2025
A newly published Apple Machine Learning Research study has challenged the prevailing idea that large-language models (LLMs) like OpenAI's o1 and Claude's thinking variants truly possess "reasoning" capabilities. The study indicates fundamental limitations in these AI systems. For this study, Apple researchers designed controllable puzzle environments, such as the Tower of Hanoi and the River Crossing. This approach avoided standard math benchmarks, which are susceptible to data contamination. According to the researchers, these custom environments allowed for a precise analysis of both the final answers produced by the LLMs and their internal reasoning traces across different complexity levels. What Apple researchers have found out from this study According to a report by MacRumors, the reasoning models tested by Apple's Research team, including o3-mini, DeepSeek-R1, and Claude 3.7 Sonnet, saw their accuracy collapse entirely once problem complexity crossed certain thresholds. Success rates dropped to zero even though the models had sufficient computational resources. Surprisingly, as problems became harder, the models reduced their reasoning effort. This points to fundamental scaling limitations rather than a lack of resources. Even more revealing, the models still failed at the same complexity points even when researchers provided complete solution algorithms. This indicates that the limitation lies in basic logical step execution, not in choosing the right problem-solving strategy. The models also showed puzzling inconsistencies. They were able to solve problems requiring over 100 moves but failed on simpler puzzles that needed only 11 moves. The study identified three performance patterns. Standard models unexpectedly performed better than reasoning models on low-complexity problems. Reasoning models had an advantage at medium complexity. Both types failed at high complexity. Researchers also discovered that models exhibited inefficient "overthinking" patterns, often discovering correct solutions early but wasting computational effort exploring incorrect alternatives. The key takeaway is that current "reasoning" models rely heavily on advanced pattern matching, not true reasoning. These models do not scale their reasoning the way humans do. They tend to overthink easy problems and think less when faced with harder ones. It is worth noting that this research surfaced just days before WWDC 2025. According to Bloomberg, Apple is expected to focus on new software designs rather than headline-grabbing AI features at this year's event. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
