Google Deep Mind's Chief Executive Officer Demmis Hassabis announced 'Gemma 4', the company's latest open artificial intelligence (AI) model that specialises in agentic workflows and advanced reasoning, via a post on X on Thursday.
Open models are designed for programmers to take apart, modify and tailor to execute specific functions.
Hassabis said in his post that Gemma 4 comes in four sizes that can be fine-tuned for specific tasks.
The first one being 31B, which was described as "dense for great raw performance". This means it has a high level of accuracy and gives out comprehensive and intelligent output but requires powerful computers to run.
The second one, 26B MoE, is for low latency, which responds faster and is more efficient, but may give out comparatively less accurate and intelligent output.
ALSO READ: Google's TurboQuant For Efficient AI Systems Triggers Selloff In SanDisk, Micron, Western Digital
The third and final ones are 2B and 4B for edge device use cases, which means that they can run on phones and other lightweight devices, and are less powerful.
Google CEO Sundar Picchai reposted the announcement and stated that Gemma 4 is "packing an incredible amount of intelligence per parameter."
"Gemma 4 enables multi-step planning, autonomous action, offline code generation, and even audio-visual processing, all without specialized fine-tuning. It's also built for a global audience with support for over 140 languages," the company said in its blog post.
Users can sort through complex workflows and create their own applications entirely form conversations with Gemma 4, the blog said.
ALSO READ: Google Introduces TranslateGemma, Intensifying Rivalry With ChatGPT Translate
Essential Business Intelligence, Continuous LIVE TV, Sharp Market Insights, Practical Personal Finance Advice and Latest Stories — On NDTV Profit.