Gemma 4 is here: All you need to know about Google’s new open AI models – Storyboard18
ADVERTISEMENT
Google is doubling down on open AI and this time, it is going smaller and smarter. The company has launched Gemma 4, its latest open model family, designed to bring powerful AI capabilities to everything from data centre systems to smartphones.
The launch builds on significant developer adoption of earlier versions. Google said Gemma models have been downloaded over 400 million times, leading to the creation of more than 100,000 variants within its developer ecosystem.
Google Chief Executive Officer Sundar Pichai said the new models deliver an “incredible amount of intelligence per parameter.” Google DeepMind Chief Executive Officer Demis Hassabis described Gemma 4 as the “best open models in the world for their respective sizes,” highlighting improvements in efficiency and performance.
Four model sizes across use cases
Gemma 4 has been released in four configurations to address different computing needs. These include Effective 2B and Effective 4B models designed for mobile and edge devices, alongside a 26 billion parameter Mixture of Experts model and a 31 billion parameter dense model aimed at higher-end systems.
Google said the 31B model is currently ranked among the top open models globally on the Arena AI leaderboard, with performance that outpaces models significantly larger in size.
Focus on reasoning, agents and multimodal capabilities
According to the company, Gemma 4 expands beyond basic conversational tasks to support advanced reasoning, multi-step problem solving and improved instruction handling. It also enables agent-based workflows through features such as function calling, structured outputs and system-level instructions.
The models support offline code generation, allowing developers to run them locally as coding assistants. They are also multimodal, with the ability to process text, images and video, while smaller variants additionally support audio input for speech-related tasks.
Gemma 4 offers extended context windows of up to 256,000 tokens for larger models and 128,000 tokens for edge variants, along with native training across more than 140 languages.
Built for mobile and offline use
Google emphasised the efficiency of the smaller models, which were developed in collaboration with its Pixel team as well as chipmakers Qualcomm and MediaTek. These versions are designed to run fully offline with minimal latency on devices such as smartphones, Raspberry Pi systems and Nvidia Jetson hardware.
The company added that Gemma 4 is built on the same research foundation as its Gemini 3 models, bringing similar capabilities to an open and customisable framework.
Gemma 4 is being released under the Apache 2.0 licence, allowing developers and organisations to freely use, modify and deploy the models across environments. Google said this approach is intended to provide flexibility and encourage broader innovation in AI development.
ADVERTISEMENT
As the creator economy explodes, influencer marketing agencies are emerging as the real power brokers reshaping how corporate India spends its advertising money.
ADVERTISEMENT
A night of ideas, impact and recognition—Storyboard18 Awards for Creativity will unveil its winners on April 7.
ADVERTISEMENT
Brand Makers
HUL names three CMOs as ‘Unified India’ strategy reshapes marketing structure
Brand Makers
CFOs becoming CEOs: The leadership trend reshaping corporate India
Agency News
WPP Media builds new model to capture Rs 10,000 cr MSME Ad surge: Ashwin Padmanabhan, COO, WPP Media
Brand Makers
The Great FMCG Reset: What the CXO churn is really signalling
How it Works
WPP, Havas, Omnicom: Are advertising’s biggest holdcos recasting agencies as AI Operating Systems?
Digital
Delhi HC directs Meta, Google, Amazon to take down deepfake content linked to Gautam Gambhir
Brand Marketing
You can now create 3-minute-long AI music with Google’s Lyria 3 Pro
Digital
A walk through iconic Big Tech logos: Here’s how simplicity took over
Brand Makers
Why Jasjeet Sekhon’s appointment matters for Google DeepMind’s AGI push?
Digital
EU broadcasters urge DMA rules for Google, Amazon, Apple smart TVs and voice assistants
Have a query? Got feedback? Want to share tips or ideas? Our team would be happy to hear from you. Get in touch with us here:
Storyboard18@nw18.com
source
This article was autogenerated from a news feed from CDO TIMES selected high quality news and research sources. There was no editorial review conducted beyond that by CDO TIMES staff. Need help with any of the topics in our articles? Schedule your free CDO TIMES Tech Navigator call today to stay ahead of the curve and gain insider advantages to propel your business!
Google is doubling down on open AI and this time, it is going smaller and smarter. The company has launched Gemma 4, its latest open model family, designed to bring powerful AI capabilities to everything from data centre systems to smartphones.
The launch builds on significant developer adoption of earlier versions. Google said Gemma models have been downloaded over 400 million times, leading to the creation of more than 100,000 variants within its developer ecosystem.
Google Chief Executive Officer Sundar Pichai said the new models deliver an “incredible amount of intelligence per parameter.” Google DeepMind Chief Executive Officer Demis Hassabis described Gemma 4 as the “best open models in the world for their respective sizes,” highlighting improvements in efficiency and performance.
Four model sizes across use cases
Gemma 4 has been released in four configurations to address different computing needs. These include Effective 2B and Effective 4B models designed for mobile and edge devices, alongside a 26 billion parameter Mixture of Experts model and a 31 billion parameter dense model aimed at higher-end systems.
Google said the 31B model is currently ranked among the top open models globally on the Arena AI leaderboard, with performance that outpaces models significantly larger in size.
Focus on reasoning, agents and multimodal capabilities
According to the company, Gemma 4 expands beyond basic conversational tasks to support advanced reasoning, multi-step problem solving and improved instruction handling. It also enables agent-based workflows through features such as function calling, structured outputs and system-level instructions.
The models support offline code generation, allowing developers to run them locally as coding assistants. They are also multimodal, with the ability to process text, images and video, while smaller variants additionally support audio input for speech-related tasks.
Gemma 4 offers extended context windows of up to 256,000 tokens for larger models and 128,000 tokens for edge variants, along with native training across more than 140 languages.
Built for mobile and offline use
Google emphasised the efficiency of the smaller models, which were developed in collaboration with its Pixel team as well as chipmakers Qualcomm and MediaTek. These versions are designed to run fully offline with minimal latency on devices such as smartphones, Raspberry Pi systems and Nvidia Jetson hardware.
The company added that Gemma 4 is built on the same research foundation as its Gemini 3 models, bringing similar capabilities to an open and customisable framework.
Gemma 4 is being released under the Apache 2.0 licence, allowing developers and organisations to freely use, modify and deploy the models across environments. Google said this approach is intended to provide flexibility and encourage broader innovation in AI development.
ADVERTISEMENT
As the creator economy explodes, influencer marketing agencies are emerging as the real power brokers reshaping how corporate India spends its advertising money.
ADVERTISEMENT
A night of ideas, impact and recognition—Storyboard18 Awards for Creativity will unveil its winners on April 7.
ADVERTISEMENT
Brand Makers
HUL names three CMOs as ‘Unified India’ strategy reshapes marketing structure
Brand Makers
CFOs becoming CEOs: The leadership trend reshaping corporate India
Agency News
WPP Media builds new model to capture Rs 10,000 cr MSME Ad surge: Ashwin Padmanabhan, COO, WPP Media
Brand Makers
The Great FMCG Reset: What the CXO churn is really signalling
How it Works
WPP, Havas, Omnicom: Are advertising’s biggest holdcos recasting agencies as AI Operating Systems?
Digital
Delhi HC directs Meta, Google, Amazon to take down deepfake content linked to Gautam Gambhir
Brand Marketing
You can now create 3-minute-long AI music with Google’s Lyria 3 Pro
Digital
A walk through iconic Big Tech logos: Here’s how simplicity took over
Brand Makers
Why Jasjeet Sekhon’s appointment matters for Google DeepMind’s AGI push?
Digital
EU broadcasters urge DMA rules for Google, Amazon, Apple smart TVs and voice assistants
Have a query? Got feedback? Want to share tips or ideas? Our team would be happy to hear from you. Get in touch with us here:
Storyboard18@nw18.com
source
This article was autogenerated from a news feed from CDO TIMES selected high quality news and research sources. There was no editorial review conducted beyond that by CDO TIMES staff. Need help with any of the topics in our articles? Schedule your free CDO TIMES Tech Navigator call today to stay ahead of the curve and gain insider advantages to propel your business!
