Google unpacks Gemma 4 capable of multi-step planning, deep logic
U.S. IT major, Google, Thursday April 02, 2026 unpacked Gemma 4 - a new generation of open models capable of multi-step planning and deep logic
U.S. IT major, Google, Thursday April 02, 2026 unpacked Gemma 4 - a new generation of open models capable of multi-step planning and deep logic.
Describing it as its most intelligent open model family to date, Google said Gemma 4 is built on the same research and technology foundation as Gemini 3 and is released under the Apache license 2.0.
"They complement our Gemini models, giving developers the industry's most powerful combination of both open and proprietary tools", Google said.
In a blog post, Google said the new family is intended to make advanced artificial intelligence capabilities more widely accessible to developers.
Google said Gemma 4 is being released in four different sizes - Effective 2B (E2B), Effective 4B (E4B), 26B Mixture of Experts (MoE) and 31B Dense, to support a range of hardware from Android devices and laptop GPUs to developer workstations and accelerators.
"The entire family moves beyond simple chat to handle complex logic and agentic workflows", Google said.
Gemma 4 can be used for building autonomous agents that interact with tools and application programming interfaces, Google said, adding that the four Gemma 4 models also support code generation and can natively process images and video.
According to Google, the edge models feature a 128K context window, while the larger models support up to 256K context. The company also said Gemma 4 was natively trained on more than 140 languages.
Google said Gemma has been downloaded more than 400 million times since the launch of its first generation, and that the "Gemmaverse" community has created more than 100,000 variants.
Follow ummid.com WhatsApp Channel for all the latest updates.
Select Language to Translate in Urdu, Hindi, Marathi or Arabic
