Gemma 4 12B StyleTune — Hardware Requirements & GPU Compatibility
ChatRoleplay
Gemma 4 12B StyleTune is a 13.0B-parameter open language model from Gryphe in the Gemma 4 family. It supports a context window of up to 262,144 tokens. At BF16 it needs about 26.99 GB of VRAM — see which GPUs and Macs can run it below.
746 downloads
25 likesSpecifications
- Publisher
- Gryphe
- Family
- Gemma 4
- Parameters
- 13.0B
- Architecture
- Gemma4UnifiedForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 262,144
- Release Date
- 2026-06-19
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Gemma 4 12B StyleTune Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16est. | 16.00 | 27.0 GB | 122.9 GB | 25.93 GB | Brain floating point 16 — preferred for training |
est.= calculated VRAM estimate; no published GGUF file found for that quantization yet. Other rows are verified against real community uploads.
Which GPUs Can Run Gemma 4 12B StyleTune?
BF16 · 27.0 GBGemma 4 12B StyleTune (BF16) requires 27.0 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 36+ GB is recommended. Using the full 262K context window can add up to 95.9 GB, bringing total usage to 122.9 GB. 1 GPU can run it, including NVIDIA GeForce RTX 5090.
All compatible consumer-level GPUs are running near their VRAM limit. You may also want to consider professional GPUs (e.g., NVIDIA A100, H100) which offer significantly more VRAM. For more headroom and better throughput, consider a multi-GPU configuration with tensor parallelism (supported by tools like vLLM, llama.cpp, or text-generation-inference).
Decent
— Enough VRAM, may be tightWhich Devices Can Run Gemma 4 12B StyleTune?
BF16 · 27.0 GB31 devices with unified memory can run Gemma 4 12B StyleTune, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (36 GB).
Runs great
— Plenty of headroomDecent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does Gemma 4 12B StyleTune need?
Gemma 4 12B StyleTune requires 27.0 GB of VRAM at BF16. Full 262K context adds up to 95.9 GB (122.9 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 13.0B × 16 bits ÷ 8 = 25.9 GB
KV Cache + Overhead ≈ 1.1 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 97 GB (at full 262K context)
VRAM usage by quantization
BF1627.0 GBBF16 + full context122.9 GB- Can I run Gemma 4 12B StyleTune on a Mac?
Gemma 4 12B StyleTune requires at least 27.0 GB at BF16, 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 12B StyleTune locally?
Yes — Gemma 4 12B StyleTune can run locally on consumer hardware. At BF16 quantization it needs 27.0 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Gemma 4 12B StyleTune?
At BF16, Gemma 4 12B StyleTune can reach ~163 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 27.0 × 0.65 = ~193 tok/s
Estimated speed at BF16 (27.0 GB)
~193 tok/s~193 tok/s~163 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 12B StyleTune?
At BF16, the download is about 25.93 GB.
- Which GPUs can run Gemma 4 12B StyleTune?
1 consumer GPU can run Gemma 4 12B StyleTune at BF16 (27.0 GB). Top options include NVIDIA GeForce RTX 5090.
- Which devices can run Gemma 4 12B StyleTune?
35 devices with unified memory can run Gemma 4 12B StyleTune at BF16 (27.0 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.