The United States has repatriated a Chinese national suspected of drug smuggling and trafficking to China. This marks a significant development in bilateral cooperation on narcotics control. The handover occurred based on information provided by China's anti-drug authorities
The Union Parliamentary Affairs Minister Kiren Rijiju on Thursday said that the Budget Session of Parliament will be reconvened on April 16 for three days to introduce “important bills”, ANI reported. Reports indicated that the extension to the ongoing session, which was to end on Thursday
Kolkata Knight Riders captain Ajinkya Rahane, usually composed during media interactions, delivered a blunt response to those questioning his batting style after his team’s 65-run loss to Sunrisers Hyderabad in the IPL on Thursday, April 2. Kolkata Knight Riders captain Ajinkya Rahane
Delhi-NCR witnessed a dust storm on Friday morning. IMD has predicted thunderstorms, rain and strong winds. AQI hit 218 in “poor” category. Residents of Delhi and its adjoining areas, including Noida, woke up to a dust storm and hazy conditions as a sudden shift in weather swept across the
The 'Maamla Legal Hai' actor stressed on the importance of gaining recognition as an actor for the content creators to receive good opportunities in the entertainment industry. New Delhi: Actress Kusha Kapila called the common critical notion of casting content creators in the OTT series and movies
Gemma 4 imbibes Google’s sharpest AI instincts, and is more welcoming
Gemma 4 imbibes Google’s sharpest AI instincts, and is more welcoming
Gemma 4 imbibes Google’s sharpest AI instincts, and is more welcoming
Google’s Gemma 4 open models make a strong case for an open AI that can run locally, with an eye on competition from China
Published on: Apr 03, 2026 9:13 AM IST By Vishal Mathur Share via Copy link Even as Google releases the Gemma 4, its open large language models, there is a short historical significance to it. These models find the same technological and research foundations as the Gemini 3 Pro models, released late last year, and marked a big step forward for proprietary large language models. Gemma 4 brings those improvements for open models, with a focus on advanced reasoning and agentic workflows. It is natively trained on over 140 languages. The convenience and flexibility extend across personal, commercial, and enterprise usage (Official image) Gemma 4, with four being the generation reference in the name, is available in four model sizes as well. There’s the Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts (MoE), and 31B Dense. The 26B and 31B models are geared for frontier intelligence, with offline compute on personal systems. The smaller parameter size E2B and E4B models are more geared for smartphones, mobile devices, and internet of things (IoT) ecosystems, as well as edge devices, including the Raspberry Pi, and Nvidia’s Jetson. In case you are wondering how Gemma is different from Gemini, the key difference is that Gemma is an AI processing engine, and not a chatbot-esque implementation. While the underlying technology is largely similar, Gemma is an open model that can be downloaded and run locally for free. It also brings much more flexibility to modify, fine-tune, and customise, depending on specific workflows and computing needs. The fact that these models can be run locally and not on the cloud also has the benefits of data privacy and cost, for many applications. The convenience and flexibility extend across personal, commercial, and enterprise usage. Google has worked closely with partners, including Qualcomm, MediaTek, and Nvidia, for these models. The new models are released under the Apache 2.0 licence, which means offering significant freedom for developers, researchers, and commercial entities to use, modify, fine-tune, and redistribute the models with minimal restrictions. Previously, that flexibility was comparatively limited. “Purpose-built for advanced reasoning and agentic workflows, Gemma 4 delivers an unprecedented level of intelligence-per-parameter. This breakthrough builds on incredible community momentum: since the launch of our first generation, developers have downloaded Gemma over 400 million times, building a vibrant Gemmaverse of more than 100,000 variants,” says Clement Farabet, Vice President of Research at Google DeepMind. Arena AI, a public, web-based platform that evaluates large language models (LLMs), ranks Gemma 4’s 31 billion parameter model at number three (score of 1452), behind GLM-5 (1456 score) and Kimi K2.5 Thinking 1453 score), while the 26 billion parameter model ranks sixth (1441 score). GLM-5 is made by the Chinese AI company Z.ai, while the Kimi models are developed by the Chinese company Moonshot AI. OpenAI’s gpt-oss-20b open-weight language model, released in August, scores 1318. Farabet notes that Gemma 4 outcompetes models 20x its size. Gemma 4 is expected to solve a wide variety of generative AI tasks with text, audio, and image input, support for over 140 languages, and a long 128K and up to 256K context window. The 31B and 26B parameter models are designed for higher-end servers with powerful GPUs, including Nvidia’s H100. “The 26B and 31B models are designed for high-performance reasoning and developer-centric workflows, making them well-suited for agentic AI. Optimized to deliver state-of-the-art, accessible reasoning, these models run efficiently on NVIDIA RTX GPUs and DGX Spark — powering development environments, coding assistants, and agent-driven workflows,” notes Michael Fukuyama, Product Manager at Nvidia. The 2 billion and 4 billion parameter footprint of the E2B and E4B models will be crucial for Gemma 4’s utility on mobile devices, with optimisations to preserve RAM and battery life. Nvidia confirms that the Jetson Orin Nano supports the Gemma 4 e2b and e4b variants, enabling multimodal inference on small, embedded, and power-constrained systems, with the same model family scaling across the Jetson platform up to Jetson Thor. ABOUT THE AUTHOR Vishal Mathur Vishal Mathur is Technology Editor for Hindustan Times. When not making sense of technology, he often searches for an elusive analog space in a digital world. Stay updated with the latest Business News on Petrol Price, Gold Rate, Silver Rates, Diesel Prices along with Income Tax Calculator Home / Business / Gemma 4 Imbibes Google’s Sharpest AI Instincts, And Is More Welcoming See LessEven as Google releases the Gemma 4, its open large language models, there is a short historical significance to it. These models find the same technological and research foundations as the Gemini 3 Pro models, released late last year, and marked a big step forward for proprietary large language models. Gemma 4 brings those improvements for open models, with a focus on advanced reasoning and agentic workflows. It is natively trained on over 140 languages.
Gemma 4, with four being the generation reference in the name, is available in four model sizes as well. There’s the Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts (MoE), and 31B Dense. The 26B and 31B models are geared for frontier intelligence, with offline compute on personal systems. The smaller parameter size E2B and E4B models are more geared for smartphones, mobile devices, and internet of things (IoT) ecosystems, as well as edge devices, including the Raspberry Pi, and Nvidia’s Jetson.
In case you are wondering how Gemma is different from Gemini, the key difference is that Gemma is an AI processing engine, and not a chatbot-esque implementation. While the underlying technology is largely similar, Gemma is an open model that can be downloaded and run locally for free. It also brings much more flexibility to modify, fine-tune, and customise, depending on specific workflows and computing needs. The fact that these models can be run locally and not on the cloud also has the benefits of data privacy and cost, for many applications.
The convenience and flexibility extend across personal, commercial, and enterprise usage.
Google has worked closely with partners, including Qualcomm, MediaTek, and Nvidia, for these models. The new models are released under the Apache 2.0 licence, which means offering significant freedom for developers, researchers, and commercial entities to use, modify, fine-tune, and redistribute the models with minimal restrictions. Previously, that flexibility was comparatively limited.
“Purpose-built for advanced reasoning and agentic workflows, Gemma 4 delivers an unprecedented level of intelligence-per-parameter. This breakthrough builds on incredible community momentum: since the launch of our first generation, developers have downloaded Gemma over 400 million times, building a vibrant Gemmaverse of more than 100,000 variants,” says Clement Farabet, Vice President of Research at Google DeepMind.
Arena AI, a public, web-based platform that evaluates large language models (LLMs), ranks Gemma 4’s 31 billion parameter model at number three (score of 1452), behind GLM-5 (1456 score) and Kimi K2.5 Thinking 1453 score), while the 26 billion parameter model ranks sixth (1441 score). GLM-5 is made by the Chinese AI company Z.ai, while the Kimi models are developed by the Chinese company Moonshot AI. OpenAI’s gpt-oss-20b open-weight language model, released in August, scores 1318.
Farabet notes that Gemma 4 outcompetes models 20x its size. Gemma 4 is expected to solve a wide variety of generative AI tasks with text, audio, and image input, support for over 140 languages, and a long 128K and up to 256K context window. The 31B and 26B parameter models are designed for higher-end servers with powerful GPUs, including Nvidia’s H100.
“The 26B and 31B models are designed for high-performance reasoning and developer-centric workflows, making them well-suited for agentic AI. Optimized to deliver state-of-the-art, accessible reasoning, these models run efficiently on NVIDIA RTX GPUs and DGX Spark — powering development environments, coding assistants, and agent-driven workflows,” notes Michael Fukuyama, Product Manager at Nvidia.
The 2 billion and 4 billion parameter footprint of the E2B and E4B models will be crucial for Gemma 4’s utility on mobile devices, with optimisations to preserve RAM and battery life. Nvidia confirms that the Jetson Orin Nano supports the Gemma 4 e2b and e4b variants, enabling multimodal inference on small, embedded, and power-constrained systems, with the same model family scaling across the Jetson platform up to Jetson Thor.
Source: HindustanTimes
Related Posts: Dollar makes a soft start to 2026 after sharpest drop in 8 years Younger Canadians and housing crisis drive sharpest anti-immigration shift in 50 years Canada Records Sharpest Population Decline In 80 Years Google tweaks Google Play conditions following EU pressure Google Rolls Out New AI-Powered Editing Feature For Google Photos- Details Google Rolls Out New AI-Powered Editing Feature For Google Photos- Details ChatGPT wants to take on Google but with Google's own tools Google engineer says Claude built in 1 hour what it took Google 1 year to do Bengaluru techie quits Google to follow his dreams Andhra Pradesh clinches $10 billion Google investment
Kerala once made history with the world’s first democratically elected Communist government. Sixty-seven years later, it is the Left’s only ruling state in India. Meanwhile, it vanished from West Bengal after 34 years in power and lost even Opposition status in Tripura
2 minutes ago
NSE, BSE Holiday List 2026: India is observing Good Friday today, April 3, a key Christian festival marked by solemn observances across the country. According to the official holiday calendar, trading on the National Stock Exchange (NSE) and the BSE remains shut for the day
2 minutes ago
I set out from Visakhapatnam with the misplaced confidence of someone who has only admired adventure sports from a comfortable distance. The road to Rayagada, in southern Odisha, carries no sense of urgency. It moves through farm fields, past Bobbili and Parvathipuram
2 minutes ago
Chadha in his message to the ‘aam aadmi’, said that whenever he got a chance to speak in the Parliament, he utilised the opportunity to raise the issues of the common people. Hours after the Aam Aadmi Party (AAP) moved to replace MP Raghav Chadha as the party’s deputy leader in the Rajya
2 minutes ago
For a format born out of evasion, the speakeasy has found a surprisingly stable home in Delhi. This is a city that has never been uneasy with discretion; if anything, it has long understood that some pleasures are heightened by selective access
2 minutes ago
These commitments are expected to rise by around 45 per cent this year, potentially crossing $180 billion. New Delhi: India’s data centre (DC) capacity is projected to grow by around 30 per cent year-on-year in 2026, driven by strong demand and continued investor interest
2 minutes ago
Zeenat Aman recalls struggling to push Amitabh Bachchan in wheelchair in Don: ‘It was awfully difficult'Zeenat Aman reflected on her experience playing Roma in Don, recalling the challenges of a specific scene with Amitabh Bachchan. Apr 3, 2026
2 minutes ago
Tech Tonic - HDMI, USB-C and history repeating itselfWe survived the HDMI mess, hoped to do better, and then rebuilt the exact same complication with a smaller connector. Published on: Apr 03, 2026 9:29 AM IST By Vishal Mathur Share via Copy link It’s actually great that USB-C cables are
2 minutes ago
Khunti (Jharkhand), Apr 3 (PTI) Three children, including two brothers, allegedly drowned while taking a bath in a pond in Jharkhand’s Khunti district, a senior police officer said on Friday. The incident took place in Gafu Badka Toli village in Torpa police station area on Thursday evening
2 minutes ago
The Delhi High Court ordered the immediate takedown of the song 'Volume 1' by rappers Yo Yo Honey Singh and Badshah from all digital platforms. Mumbai: The Delhi High Court has ordered the removal of rappers Yo Yo Honey Singh and Badshah's controversial song ‘Volume 1’ from all social media and
2 minutes ago
In the 1960s Ed Dwight was the first Black astronaut candidate — but he never got his chance to go to the Moon. He said he's now living out that once-denied dream vicariously through Victor Glover, who is set to make history on the Artemis 2 Moon mission that could take off as soon as Wednesday
2 minutes ago
2026 Suzuki Burgman Street 125: Suzuki Motorcycle India has introduced the second-generation Burgman Street 125 in the domestic market, bringing a range of updates in design, features and underpinnings. 2026 Suzuki Burgman Street 125: Suzuki Motorcycle India has introduced the second-generation
2 minutes ago
United States Army chief Randy George on Thursday retired with immediate effect amid the conflict in West Asia. CBS News quoted unidentified officials as saying that Defense Secretary Pete Hegseth had asked George to step down because he wants an officer who will implement his and President Donald
2 minutes ago
Raghav Chadha breaks silence after AAP drops him from Rajya Sabha post: ‘Have I done something wrong?’Raghav Chadha breaks silence after AAP drops him from Rajya Sabha Updated on: Apr 03, 2026 10:55 AM IST By HT News Desk Share via Copy link AAP MP Raghav Chadha on Friday issued his first
2 minutes ago
Commercial LPG Supply Boost: The Uttarakhand government has rolled out a revised Standard Operating Procedure (SOP) for commercial LPG distribution to manage rising demand and ensure supply stability, replacing the earlier guidelines.Commercial LPG Supply Boost: The Uttarakhand government has
2 minutes ago
Gemma 4 imbibes Google’s sharpest AI instincts, and is more welcomingGoogle’s Gemma 4 open models make a strong case for an open AI that can run locally, with an eye on competition from China Published on: Apr 03, 2026 9:13 AM IST By Vishal Mathur Share via Copy link Even as Google releases the
2 minutes ago
US Defense Secretary Pete Hegseth has asked Army Chief of Staff Randy George to step down from his post, according to CBS News, the BBC's US partner. Chief Pentagon spokesperson Sean Parnell said in a statement on social media that George "will be retiring from his position as the 41st Chief of
2 minutes ago
Middle East war: Iran reports shooting down a US F-35 fighter jet over its territory. This comes as Iran also launched missiles towards Israel. The US Central Command has not yet confirmed the F-35 incident. Earlier, three US F-15E aircraft were downed in Kuwait due to friendly fire during
2 minutes ago
Google CEO Sundar Pichai set to return to Stanford University as a commencement speaker: ‘It’s a privilege’The CEO has a master’s degree in materials science and engineering from Stanford University. Updated on: Apr 03, 2026 8:27 AM IST By Sakshi Sah Share via Copy link Google CEO Sundar
2 minutes ago
The SRH bowling unit was under scrutiny after their first Indian Premier League (IPL) 2026 match against the defending champions Royal Challengers Bengaluru (RCB). Chasing a formidable target of 202, RCB reached the finish line with 26 balls to spare
2 minutes ago
Published - April 01, 2026 11:20 am IST The Hindu Quizzes Published - April 01, 2026 11:20 am IST The Hindu Quizzes
2 minutes ago
Editorial independence is core to our work. Some links may earn us a commission, without influencing our opinions.No more ‘Can you repeat that?’: Top noise-cancelling earphones for better office callsNoisy office environments can hinder calls. These earphones can improve clarity during calls
2 minutes ago
A special session of Parliament will be convened for three days beginning April 16 to pass key Constitution amendment bills aimed at increasing the number of Lok Sabha seats from 543 to 816, enabling the implementation of the women’s reservation law at the earliest
2 minutes ago
The four astronauts on the Artemis II mission have now left the Earth's orbit, after their Orion spacecraft fired its main engine for a final push towards the Moon. The five minute and 55 second engine burn, known as the translunar injection (TLI), went "flawlessly"
2 minutes ago
iQOO Z11x 5G Review: The smartphone delivers strong performance, a big 7200mAh battery, and solid durability at a budget price. It runs on Android 16 with long-term updates and handles gaming smoothly, though the camera is average and the LCD display is not the best
2 minutes ago
The Directorate of Education (DoE) has issued one of its strictest directives yet, aiming to curb the exploitation of parents by private schools in Delhi, particularly with the new academic session approaching. The administration has made it clear that education cannot be treated as a commercial
2 minutes ago
Having reached the World Group playoffs for only the second time last year, the Indian women’s tennis team would be hoping to repeat the feat while playing at home for the second successive time when the Billie Jean King Cup Asia/Oceania Group I gets underway here from April 7
2 minutes ago
