Running Gemma 4 for free on Kaggle: shrinking it from 62 GB to 18 GB
Google Gemma 4 31B takes up about 62 GB in float16, yet the author managed to run it on free Kaggle, which caps disk space at 57.6 GB. The trick: quantize the model to NF4 on the fly, split it across two T4 GPUs with device_map="auto", delete the Hugging Face cache once loading finishes, and push the result straight to the Hub. The final model weighs 18.3 GB.
AI-processed from Habr AI; edited by Hamidun News
Google Gemma 4, a large 31-billion-parameter model, has been squeezed from 62 GB down to 18.3 GB and run for free on Kaggle, even though the platform's disk limit is smaller than the model's original size.
Where can you run Google Gemma 4 for free?
On Kaggle, by quantizing the model down to 18.3 GB. The catch is that Kaggle's disk is capped at 57.6 GB, while the original model weighs 62 GB. On-the-fly quantization and uploading to the Hugging Face Hub solve the problem.
How do you shrink Google Gemma from 62 to 18.3 GB?
Quantize it to 4 bits, delete the cache automatically during execution, and upload the result directly to the Hugging Face Hub.
Can you run the model for free on Kaggle?
Yes. Google Gemma 4 was shrunk from 62 GB to 18.3 GB and run for free on Kaggle, despite the platform's 57.6 GB disk space limit.
- Nov 1, 2026Google's StreamVibe video agent beats other models by 10.7% on response timing and quality
- Nov 1, 2026OpenAI Rewrites GPT Behavior Rules: Five Values and a Chain of Command
- Nov 1, 2026Voice bots built on LLMs turn to Jev, a fast classifier, for decisions in 200–500 ms
- Nov 1, 2026Google admits AI Mode cuts search revenue, shares fall 15%
Key Questions Answered
Where can you run Google Gemma 4 for free?
On Kaggle, by quantizing the model down to 18.3 GB. The catch is that Kaggle's disk is capped at 57.6 GB, while the original model weighs 62 GB. On-the-fly quantization and uploading to the Hugging Face Hub solve the problem.
How do you shrink Google Gemma from 62 to 18.3 GB?
Quantize it to 4 bits, delete the cache automatically during execution, and upload the result directly to the Hugging Face Hub.
Can you run the model for free on Kaggle?
Yes. Google Gemma 4 was shrunk from 62 GB to 18.3 GB and run for free on Kaggle, despite the platform's 57.6 GB disk space limit.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.