Google releases Gemma 4 12B: 150 million downloads, runs on a laptop with 16 GB of VRAM

Google DeepMind has released Gemma 4 12B, a new version of its open Gemma language model series, and at the same time announced a record 150 million total downloads across the whole lineup. The model is compact enough to run locally on a laptop with 16 GB of VRAM, and it is distributed under the Apache 2.0 license.

A model that fits on a laptop

The main practical feature of Gemma 4 12B is its modest hardware requirements. 16 gigabytes of video memory is enough to run it, which matches the specs of common consumer GPUs in the RTX 4070/4080 class and their mobile versions. That lets developers work with the model without investing in specialized server GPUs.

Demis Hassabis, CEO of Google DeepMind, described Gemma 4 12B as "incredibly powerful for such a small size." According to the team, the 12-billion-parameter model retains the quality expected from much larger systems, without requiring cloud infrastructure. That makes it especially attractive for tasks where data privacy, offline operation or minimal latency matter.

For businesses, running locally means lower operating costs: one-time compute resources on their own hardware instead of constant API calls. For researchers and independent developers, it means full control over their infrastructure with no subscription fee.

What the release includes

Alongside Gemma 4 12B, Google confirmed the model's key characteristics:

• 12 billion parameters: enough for writing code, analyzing text and handling complex dialogues
• 16 GB VRAM: runs on common work and gaming laptops without modification
• Apache 2.0: full freedom for commercial use, fine-tuning and distribution
• Support for Ollama, llama.cpp, Hugging Face: integration with popular local deployment frameworks

The record 150 million downloads of the Gemma 4 series as a whole is one of the largest figures among open models. It is comparable in scale to the audiences of Meta's Llama* and Mistral, and confirms that Google has firmly established itself in the open-weights segment.

Apache 2.0 as a strategic choice

The Apache 2.0 license is not just a legal detail but a competitive decision. It allows the model to be embedded in commercial products, fine-tuned on corporate data, and derivative works to be distributed without any royalties or approval from Google. These are the terms that make the model attractive to enterprise developers who want to avoid dependence on cloud providers.

"Apache 2.0, happy building!"

Hassabis wrote in the release announcement.

At the same time, the open license helps Google build an ecosystem: companies that build products on Gemma are highly likely to stay on Google Cloud infrastructure for workloads that require scale. It is a proven strategy: open the base toolkit to keep users on the platform.

What this means

Gemma 4 12B meets a practical market demand: a powerful open model that runs without the cloud on a developer's standard hardware. For companies sensitive to privacy or looking to cut AI costs, it is a direct alternative to API dependence.

The record 150 million downloads confirm that the segment of local open models is growing, and Google intends to lead it.

*Meta is designated an extremist organization and banned in Russia.