Gemma 4 31B IT NVFP4 — Hardware Requirements & GPU Compatibility
Chat
Gemma 4 31B IT NVFP4 is a 20.9B-parameter open language model from NVIDIA in the Gemma family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 14.14 GB of VRAM — see which GPUs and Macs can run it below.
2.3M downloads
505 likesSpecifications
- Publisher
- NVIDIA
- Family
- Gemma
- Parameters
- 20.9B
- Architecture
- Gemma4ForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 262,144
- Release Date
- 2026-05-08
- License
- Other
Get Started
HuggingFace
How Much VRAM Does Gemma 4 31B IT NVFP4 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q4_K_M | 4.80 | 14.1 GB | 181.9 GB | 12.52 GB | 4-bit medium quantization — most popular sweet spot |
| Q6_K | 6.60 | 18.8 GB | 186.6 GB | 17.22 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 22.5 GB | 190.3 GB | 20.87 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run Gemma 4 31B IT NVFP4?
Q4_K_M · 14.1 GBGemma 4 31B IT NVFP4 (Q4_K_M) requires 14.1 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 19+ GB is recommended. Using the full 262K context window can add up to 167.8 GB, bringing total usage to 181.9 GB. 17 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 5080.
Runs great
— Plenty of headroom
Decent
— Enough VRAM, may be tightWhich Devices Can Run Gemma 4 31B IT NVFP4?
Q4_K_M · 14.1 GB27 devices with unified memory can run Gemma 4 31B IT NVFP4, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Mini M4 (16 GB).
Runs great
— Plenty of headroomRelated Models
Frequently Asked Questions
- How much VRAM does Gemma 4 31B IT NVFP4 need?
Gemma 4 31B IT NVFP4 requires 14.1 GB of VRAM at Q4_K_M, or 22.5 GB at Q8_0. Full 262K context adds up to 167.8 GB (181.9 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 20.9B × 4.8 bits ÷ 8 = 12.5 GB
KV Cache + Overhead ≈ 1.6 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 169.4 GB (at full 262K context)
VRAM usage by quantization
Q4_K_M14.1 GBQ4_K_M + full context181.9 GB- What's the best quantization for Gemma 4 31B IT NVFP4?
For Gemma 4 31B IT NVFP4, Q4_K_M (14.1 GB) offers the best balance of quality and VRAM usage. Q6_K (18.8 GB) provides better quality if you have the VRAM.
VRAM requirement by quantization
Q4_K_M ★14.1 GBQ6_K18.8 GBQ8_022.5 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run Gemma 4 31B IT NVFP4 on a Mac?
Gemma 4 31B IT NVFP4 requires at least 14.1 GB at Q4_K_M, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Gemma 4 31B IT NVFP4 locally?
Yes — Gemma 4 31B IT NVFP4 can run locally on consumer hardware. At Q4_K_M quantization it needs 14.1 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Gemma 4 31B IT NVFP4?
At Q4_K_M, Gemma 4 31B IT NVFP4 can reach ~206 tok/s on AMD Instinct MI300X. On NVIDIA GeForce RTX 4090: ~46 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 14.1 × 0.55 = ~206 tok/s
Estimated speed at Q4_K_M (14.1 GB)
~206 tok/s~46 tok/s~154 tok/s~128 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Gemma 4 31B IT NVFP4?
At Q4_K_M, the download is about 12.52 GB. The full-precision Q8_0 version is 20.87 GB.
- Which GPUs can run Gemma 4 31B IT NVFP4?
17 consumer GPUs can run Gemma 4 31B IT NVFP4 at Q4_K_M (14.1 GB). Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090, NVIDIA GeForce RTX 3090 Ti, AMD Radeon RX 6800. 5 GPUs have plenty of headroom for comfortable inference.
- Which devices can run Gemma 4 31B IT NVFP4?
27 devices with unified memory can run Gemma 4 31B IT NVFP4 at Q4_K_M (14.1 GB), including Mac Mini M4 (16 GB), Mac Mini M4 (32 GB), Mac Mini M4 Pro (24 GB), Mac Mini M4 Pro (48 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.