News / Model Launch

Google Launches Gemma 4, Its Most Capable Open Model Family

Written by Faisal Saeed

Thu Apr 09 2026

Experience every AI model combined in one powerful tool so you never have to switch tabs again.

Google Launches Gemma 4, Its Most Capable Open Model Family

Google has introduced Gemma 4, its most capable open model family to date, purpose-built for advanced reasoning and agentic workflows.

The release builds on significant community momentum. Since the launch of the first Gemma generation, developers have downloaded the models over 400 million times, producing more than 100,000 variants in what Google calls the "Gemmaverse."

The family is released in four sizes:

Effective 2B (E2B)
Effective 4B (E4B)
A 26B Mixture of Experts (MoE)
A 31B Dense model

The 31B model currently ranks third among open models on the industry-standard Arena AI text leaderboard, while the 26B MoE model holds sixth place. Google states that Gemma 4 outcompetes models up to 20 times its size on these benchmarks.

All models in the family natively support function-calling, structured JSON output, and system instructions, enabling developers to build autonomous agents that interact with external tools and APIs. Every model also natively processes video and images at variable resolutions, with particular strength in OCR and chart understanding.

Context window sizes range from 128K tokens for the edge models up to 256K tokens for the larger variants. The entire family is natively trained in over 140 languages.

The edge-focused E2B and E4B models were developed in collaboration with Google's Pixel team and hardware partners Qualcomm Technologies and MediaTek. They are capable of running fully offline with near-zero latency on devices including phones, Raspberry Pi, and NVIDIA Jetson Orin Nano.

Gemma 4 is released under a commercially permissive Apache 2.0 open-source license and is available via Google AI Studio, Hugging Face, Kaggle, and Ollama.