🚀 Medial Secures Investment on Shark Tank India - Fueling the Future of Professional Social Networking. 🔥

Startup Showcase

Premium Content

Try our Valuation Calculator →

News on Medial

DeepSeek's updated R1 AI model is more censored, test finds | TechCrunch

TechCrunch · 3m ago

DeepSeek's updated R1 AI model is more censored, test finds | TechCrunch

DeepSeek's latest AI model, R1-0528, exhibits high performance in coding, math, and general knowledge tests, nearing OpenAI's o3 model. However, it's less responsive to controversial topics, especially those censored by the Chinese government. The model significantly limits discussions on sensitive subjects, adhering to China's 2023 law restricting content that opposes state narratives. While sometimes critical of certain policies, R1-0528 often reflects the official government stance in its responses.

1

Related News

DeepSeek updates its R1 reasoning AI model, releases it on Hugging Face | TechCrunch

TechCrunch · 3m ago

DeepSeek updates its R1 reasoning AI model, releases it on Hugging Face | TechCrunch

Chinese startup DeepSeek released an updated version of its R1 reasoning AI model on the platform Hugging Face. The R1 model, licensed under MIT for commercial use, is a minor upgrade and contains configuration files and weights. With a size of 685 billion parameters, it's challenging to run on consumer-grade hardware. DeepSeek gained attention after R1's first release and has faced scrutiny from regulators concerned about national security implications.

Market was like 'Oh my gosh, AI is finished': Huang on DeepSeek

Inshorts · 6m ago

Market was like 'Oh my gosh, AI is finished': Huang on DeepSeek

NVIDIA CEO Jensen Huang said markets misinterpreted the impact of Chinese startup DeepSeek's RI AI model on the chipmaker's business. "Market responded to R1, as in, 'Oh my gosh. AI is finished'," Huang said. Markets thought computing wasn't needed anymore but it's "complete opposite", he stated. NVIDIA's m-cap fell by $600 billion after DeepSeek's launch of its model.

China's Baidu says DeepSeek success inspired open source move

Economic Times

Economic Times · 6m ago

China's Baidu says DeepSeek success inspired open source move

Baidu is embracing an open-source model for its generative AI chatbot, inspired by DeepSeek's success. Despite initially launching an AI chatbot similar to ChatGPT, Baidu's Ernie 4.0 struggled against competitors like DeepSeek's R1. CEO Robin Li announced plans to open-source Ernie 4.5, aiming for easier adoption. Baidu's quarterly revenue fell 2% due to lower advertising, despite growth in its Cloud AI unit, handling increased user queries and interactions on the Ernie platform.

OpenAI CEO Sam Altman likely to visit India on Feb 5

YourStory

YourStory · 7m ago

OpenAI CEO Sam Altman likely to visit India on Feb 5

OpenAI's CEO Sam Altman is expected to visit India on February 5, meeting top government officials amid AI competition. Chinese firm Deepseek's low-cost AI model R1 has challenged OpenAI's dominance, overtaking ChatGPT as the top-ranked app on Apple's Appstore. Altman will participate in a fireside chat in New Delhi. His visit coincides with legal challenges OpenAI faces in India and the country's plans to develop its own affordable AI model.

DeepSeek Rolls Out V3 Model Updates, Strengthen Programming Capabilities to Outpace OpenAI

OutlookIndia

OutlookIndia · 5m ago

DeepSeek Rolls Out V3 Model Updates, Strengthen Programming Capabilities to Outpace OpenAI

Chinese AI startup DeepSeek has enhanced its V3 model with the V3-0324 update, improving programming capabilities and transitioning to the MIT open-source license. This strategic move aims to outpace competitors, offering a cost-effective AI innovation. The update, released on platforms like Hugging Face and GitHub, builds on DeepSeek's success with its R1 model. The V3 model leverages a mix of technologies for efficient, complex task handling in coding and mathematics.

DeepSeek may have used Google's Gemini to train its latest model | TechCrunch

TechCrunch · 3m ago

DeepSeek may have used Google's Gemini to train its latest model | TechCrunch

DeepSeek, a Chinese AI lab, released an updated R1 model excelling in math and coding, raising suspicions among AI researchers that it may have been trained using Google's Gemini AI outputs. Evidence by developer Sam Paeach suggests similarities in language and behavior to Gemini's, despite DeepSeek not disclosing its data sources. Accusations echo previous allegations of unauthorized data usage from AI models like OpenAI's ChatGPT, amidst ongoing concerns about the originality and ethics of AI training data.

Alibaba shares surge after it unveils reasoning model

Economic Times

Economic Times · 6m ago

Alibaba shares surge after it unveils reasoning model

Alibaba's shares rose over 8% after announcing a new reasoning model, QwQ-32B, which rivals DeepSeek's R1 in performance. Qwen, Alibaba’s AI unit, highlighted the model's capabilities in reasoning, coding, and problem-solving. This development coincides with the Chinese government's increased support for AI and tech industries. The model, part of Alibaba’s Qwen series, underscores China's growing AI ecosystem, gaining traction in both governmental and enterprise applications, aiming for immediate economic impact.

ChatGPT Founder Sam Altman is ‘impressed’ by DeepSeek, hints at new releases by OpenAI | Mint

Livemint

Livemint · 7m ago

ChatGPT Founder Sam Altman is ‘impressed’ by DeepSeek, hints at new releases by OpenAI | Mint

OpenAI's Sam Altman praised Chinese AI startup DeepSeek's R1 model for its cost-effective performance, acknowledging the invigorating presence of new competition. Altman hinted at OpenAI's plans to release better models while emphasizing the importance of computing power in advancing AI research. DeepSeek, founded by Liang Wenfeng, gained attention after its V3 large language model outperformed U.S. counterparts at lower costs. This challenges the notion that AI requires massive computing investments, presenting competitive concerns for Western tech giants.

‘Ready to invest $1M and 5 hours per week to make India great in AI’: Perplexity AI CEO

YourStory

YourStory · 7m ago

‘Ready to invest $1M and 5 hours per week to make India great in AI’: Perplexity AI CEO

Aravind Srinivas, CEO of Perplexity AI, committed $1 million and five weekly hours to boost AI in India, challenging DeepSeek's AI model. He promises an additional $10 million if Indian efforts surpass DeepSeek R1's benchmarks. Srinivas emphasized the importance of both building on existing AI models and training new ones, critiquing Nandan Nilekani's solely build-focused stance. Meanwhile, the US announced a $500 billion AI venture through OpenAI, SoftBank, Oracle, and MGX.

India Can Use DeepSeek Model to Leapfrog AI Progress: Ola's Bhavish Aggarwal

OutlookIndia

OutlookIndia · 7m ago

India Can Use DeepSeek Model to Leapfrog AI Progress: Ola's Bhavish Aggarwal

Bhavish Aggarwal, founder of Ola and AI startup Krutrim, highlighted the potential of adopting DeepSeek's open-source AI model in India, with caution towards the Chinese app itself. Emphasizing secure deployment on Indian servers, Aggarwal announced the global debut of DeepSeek-R1 671B on H100s, available to Indian developers at an introductory price. He also shared Krutrim's commitment to open-source AI development to advance India's AI capabilities and its practical integration within Ola's ecosystem.

Trackers

Active Indian VC’s

OG Capital Email

With a hands-on approach, OG Capital aims to invest in over 20 promising...

Accel Partners Email

Early and growth-stage investments in disruptive technology companies with...

Early-stage venture capital firm investing in technology startups in India. Focus on...

Access All Trackers

Startup Showcase Winners

Aug 2025

Build. Comply. Fund. Fly.

Crypto to INR payments app for Indians

Your CV, Done Right.

Enter Ongoing Startup Showcase

Top Users

Trending News on Medial

Download the medial app to read full posts, comements and news.