Latest news with #ArtificialAnalysisIntelligenceIndex
Economic Times
20-07-2025
- Business
- Economic Times
Musk says xAI will make kid-friendly app 'Baby Grok'
Agencies Elon Musk-led artificial intelligence company xAI is working on a kid-friendly version of its Grok chatbot, the billionaire said in a post on social media platform X. 'We're going to make Baby Grok @xAI, an app dedicated to kid-friendly content,' Musk posted on Sunday. The new version is expected to launch as a separate application tailored for younger users, though further details are yet to be announced. The announcement comes days after X, also owned by Musk, temporarily disabled new Grok mentions on the platform to address spam-related Musk hinted at a new capability that would enable Grok to generate viral videos, further expanding its creative tools. Grok 4 capabilities xAI launched its latest flagship model, Grok 4, on July 10, positioning it as a significant upgrade following criticism over antisemitic responses from earlier versions. The backlash included concerns raised by users on X and the Anti-Defamation League (ADL), after the chatbot was found praising Adolf Hitler in certain prompts. During a livestream event, Musk and the xAI team also introduced a $300-per-month AI subscription called SuperGrok Heavy, featuring the company's multi-agent version of Grok 4. The upgrade aims to improve performance across advanced reasoning and creative tasks. Benchmark performance Grok 4 recently topped the Artificial Analysis Intelligence Index, a benchmark suite assessing models across a range of capabilities: MMLU-Pro: Multitask language understanding GPQA Diamond: Graduate-level problem solving Humanity's Last Exam: General and existential reasoning LiveCodeBench: Real-time programming SciCode: Scientific computation and modeling AIME: Olympiad-level math reasoning Math-500: Advanced mathematical problem-solving Each of these benchmarks focuses on measuring the reasoning capabilities of next-generation AI models. Grok for government xAI also recently announced Grok for Government, a suite of its AI models made available for use by US federal agencies. The company said its inclusion in the General Services Administration (GSA) schedule will allow any federal department or agency to directly procure xAI the company is reportedly in talks to raise fresh capital at a potential valuation between $170 billion and $200 billion, according to the Financial Times. If finalised, this could make xAI one of the most valuable private AI companies US continues to lead the global AI race, producing 40 notable models in 2024 so far. Elevate your knowledge and leadership skills at a cost cheaper than your daily tea. What's keeping real retail investors out of the Nvidia rally If data is the new oil, are data centres the smokestacks of the digital age? The hybrid vs. EV rivalry: Why Maruti and Mahindra pull in different directions. What's best? Instagram and YouTube make billions off creators. Should they pay up for their mental health? Trent trips on the ramp. Is it still worth the splurge or time to change brands? Best way to deal with volatility, just ' Hold' for wealth creation: 7 large-cap stocks with an upside potential of up to 41% Stock picks of the week: 5 stocks with consistent score improvement with an upside potential of 16 to 38% in 1 year Headwinds, yes, but long-term story intact. 7 stocks from the engineering sector with upside potential from 21 to 42%
India Gazette
04-06-2025
- Business
- India Gazette
Chinese AI surge shapes global landscape in Q1 2025: Report
New Delhi [India], June 4 (ANI): The rise of Chinese AI emerges as one of the key factors that drove the Artificial Intelligence landscape in the first quarter of 2025, according to a report by Artificial Analysis, an independent AI benchmarking and insights provider. The US leads across reasoning models, the top 4 places on the Artificial Analysis Intelligence Index are all taken by reasoning models from US labs. However, China currently leads non-reasoning models; DeepSeek V3 0324 is the leading non-reasoning model. Models from labs outside of USA and China have continually improved, but currently don't compete for frontier intelligence. The rise of Chinese AI Labs marked a notable shift in global competition. Chinese entities released models that rivalled US counterparts, particularly in open-weight models, with DeepSeek V3 0324 leading in non-reasoning categories. The report also highlights that AI continues to progress, with the help of leading labs like OpenAI, Google, and many others, consistently advancing intelligence, efficiency, and speed. While OpenAI's o4-mini (high) maintained a lead, models like Google's Gemini 2.5 Pro and xAI's Grok 3 rapidly narrowed the gap, intensifying competition. In the field of AI, reasoning models emerged as a significant frontier. These models, which 'think' through problems by generating intermediate steps before providing an answer, became widespread across major labs. This approach has delivered substantial intelligence gains, influencing how AI processes complex tasks. Artificial Analysis also discusses the efficiency and Mixture of Experts (MoEs) in revolutionising AI inference. Costs plummeted by over 32 times since September 2024, and more than 1000x since GPT-4's 2023 launch. This was primarily due to smaller models (including MoE architectures), optimised inference, and new hardware, making advanced AI more accessible Lastly, Multimodal AI saw significant advancements. Models became more capable of handling image and audio data natively. Image generation, exemplified by OpenAI's GPT-4o, achieved new quality benchmarks, while text-to-speech models improved significantly, offering more human-like dialogue. These trends collectively define a quarter of rapid evolution, increased accessibility, and a globally competitive environment in AI. (ANI)
Time of India
04-06-2025
- Business
- Time of India
Chinese AI gains reshape global AI race in Q1 2025: Report
Chinese artificial intelligence advancements emerged as a major driving force in shaping the global AI landscape during the first quarter of 2025, according to a new report from Artificial Analysis, an independent AI benchmarking and insights provider. While the United States continues to dominate in reasoning models—with the top four spots on the Artificial Analysis Intelligence Index held by US labs—China has taken the lead in non-reasoning categories. The report notes that DeepSeek V3 0324 currently stands as the leading non-reasoning model, marking a significant milestone for Chinese AI labs, ANI reported. 'The rise of Chinese AI Labs marked a notable shift in global competition. Chinese entities released models that rivalled US counterparts, particularly in open-weight models, with DeepSeek V3 0324 leading in non-reasoning categories,' the report highlights. Labs outside the US and China have made progress but are not yet contenders in the frontier intelligence space, the report adds. The global AI landscape remains highly dynamic, with continual progress driven by key players such as OpenAI, Google, and others. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like People Born 1940-1975 With No Life Insurance Could Be Eligible For This Reassured Get Quote Undo OpenAI's o4-mini (high) has maintained its lead in overall performance, but competitors like Google's Gemini 2.5 Pro and xAI's Grok 3 have significantly narrowed the gap, intensifying the competition among top labs. Reasoning models, which solve problems by generating intermediate steps before reaching an answer, are identified in the report as a critical frontier. Their widespread adoption across major labs has yielded meaningful gains in intelligence and problem-solving capabilities. Artificial Analysis also points to technological shifts that have drastically reduced costs. The adoption of Mixture of Experts (MoE) architectures, smaller model designs, optimized inference, and new hardware have led to a more than 32-fold drop in AI inference costs since September 2024, and over 1000-fold since the launch of GPT-4 in 2023. These efficiencies are rapidly making high-performance AI more accessible. Multimodal AI has also made significant strides, particularly in image and audio processing. OpenAI's GPT-4o set new quality benchmarks in image generation, while text-to-speech capabilities across the board became increasingly lifelike, enhancing the naturalness of AI dialogue. Taken together, these trends signal a quarter marked by accelerated innovation, expanding access, and heightened global competition in the AI sector. Stay informed with the latest business news, updates on bank holidays and public holidays . AI Masterclass for Students. Upskill Young Ones Today!– Join Now
South China Morning Post
20-03-2025
- Business
- South China Morning Post
Tech war: Chinese AI models led by DeepSeek catch up with US rivals at lower prices
China's home-grown artificial intelligence (AI) models are rapidly closing their performance gap with leading US rivals while offering significantly lower prices, according to benchmarking and consultancy firm Artificial Analysis. Advertisement DeepSeek-R1, a reasoning model developed by China's most prominent AI start-up , scored 60 on the Artificial Analysis Intelligence Index, making it the third most intelligent AI model in the world, behind ChatGPT maker OpenAI's o1 and o3-mini, which scored 62 and 66, respectively. The index evaluated model performance across a variety of intelligence and reasoning data sets. In terms of pricing for developers, DeepSeek-R1 ranked eighth globally, while OpenAI's GPT-4.5 and o1 were listed as the two most expensive models. The rankings reflect how Chinese AI models are quickly catching up to their Western counterparts in model performance, while actively lowering prices amid an intense domestic price war. 05:00 Does the arrival of China's low-cost DeepSeek mean the end of Nvidia's chip dominance? Does the arrival of China's low-cost DeepSeek mean the end of Nvidia's chip dominance? 'One year ago, the AI frontier was overwhelmingly dominated by US companies,' Artificial Analysis said in a post on X last month. 'Today, nearly a dozen Chinese companies have models matching or exceeding current-generation models from most US labs.'
