Google releases Gemma 4, a family of open models built off of Gemini 3
When Google released Gemini 3 Pro at the end of last year, it was a significant step forward for the company's proprietary large language models. Now, the company is bringing some of the same technology and research that made those models possible to the open source community with the release of its new family of Gemma 4 open-weight models. Google is offering four different versions of Gemma 4, differentiated by the number of parameters on offer. For edge devices, including smartphones, the company has the 2-billion and 4-billion "Effective" models. For more powerful machines, thereās the 26-billion "Mixture of Experts" and 31-billion "Dense" systems. For the unfamiliar, parameters are the settings a large language model can tweak to generate an output. Typically, models with more parameters will deliver better answers than ones with less, but running them also requires more powerful hardware. With Gemma 4, Google claims it's managed to engineer systems with "an unprecedented level of intelligence-per-parameter." To back up this claim, the company points to the performance of Gemma 4's 31-billion and 26-billion variants, which claimed the third and sixth spots respectively on Arena AI's text leaderboard , beating out models 20 times their size. All of the models can process video and images, making them ideal for tasks like optical character recognition. The two smaller models are also capable of processing audio inputs and understanding speech. Separately, Google says the Gemma 4 family is capable of generating offline code, meaning you could use them to do vibe coding
Google has recently released a new family of open-source models called Gemma 4, built upon the foundation of its proprietary Gemini 3 models. This move represents a significant step for the company, as it brings cutting-edge technology and research to the open-source community. The Gemma 4 family consists of four different versions, each offering a distinct number of parameters tailored to various hardware capabilities.
For edge devices, such as smartphones, Google has introduced the 2-billion and 4-billion "Effective" models. These models are designed to run efficiently on devices with limited computational resources, making them suitable for everyday use. On the other hand, for more powerful machines, Google has developed the 26-billion "Mixture of Experts" and 31-billion "Dense" systems. These larger models require more powerful hardware but offer enhanced performance and capabilities.
Parameters, which are the settings a large language model can tweak to generate an output, play a crucial role in determining a model's capabilities. Generally, models with more parameters deliver better results, but they also demand more computational resources. Google claims that Gemma 4 models have achieved an "unprecedented level of intelligence-per-parameter," meaning they deliver exceptional performance relative to their size. To support this claim, the company highlights the performance of its 31-billion and 26-billion variants, which secured third and sixth places, respectively, on Arena AI's text leaderboard. Notably, these models outperformed others that were 20 times their size.
Gemma 4 models are not limited to text processing; they can also handle video and images, making them ideal for tasks such as optical character recognition. Additionally, the two smaller models in the family are capable of processing audio inputs and understanding speech, expanding their applicability in various real-world scenarios.
Separately, Google has announced that the Gemma 4 family can generate offline code, enabling users to perform tasks like "vibe coding" without an internet connection. This feature adds another layer of versatility to the models, allowing them to function in environments where connectivity is limited or unavailable.
Furthermore, Google has trained the Gemma 4 models in more than 140 languages, ensuring broad linguistic support and making them accessible to a global audience. This multilingual capability enhances the models' utility in diverse applications and regions.
Google is releasing the Gemma 4 family under the Apache 2.0 license, granting users greater freedom and flexibility in utilizing and modifying the models. Previously, the company offered its Gemma models under its own Gemma license. This shift to an open-source license is expected to foster collaboration and innovation within the developer community, as it encourages the exploration and adaptation of these advanced models.
In summary, Google's release of the Gemma 4 family marks a significant milestone in the company's efforts to democratize access to its proprietary large language models. With a range of models tailored to different hardware configurations and capabilities, Gemma 4 offers a versatile solution for various applications, from edge devices to powerful machines. The models' exceptional performance, multilingual support, and open-source licensing make them a valuable resource for developers and researchers alike.
