Gemma 3 12B Pt — Hardware Requirements & GPU Compatibility
VisionGemma 3 12B PT is Google's roughly 12-billion-parameter pretrained checkpoint in the Gemma 3 family, built on a vision-language architecture that accepts both text and images. The PT suffix marks the base model: it lacks instruction tuning and doesn't handle chat prompts well, so it serves as a starting point for fine-tuning or research instead of direct use as an assistant. Hugging Face lists this version as image-text-to-text. At this size it fits on a single consumer GPU once quantized, while the instruction-tuned variant remains the standard choice for conversation. It supports a 128K-token context window. Distribution follows the Gemma Terms of Use, and access requires acknowledging the license on Hugging Face. Released in March 2025, it falls between the 4B and 27B models within the Gemma 3 lineup.
Runs great on a 24 GB GPU
Runs from 8 GB (IQ2_M · 16K)Best on 24 GB (Q4_K_M · ~82 tok/s · 32K)Macs from 16 GB (Q4_K_M · ~17 tok/s)
Gemma Terms
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Gemma 3 12B Pt
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/mradermacher/gemma-3-12b-pt-i1-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 13.1 GB free · 16K context
Download: gemma-3-12b-pt.i1-Q4_K_M.gguf7.30 GBmradermacher
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 5070 Ti (16 GB) (5 GB headroom) · Q4_K_M
- Generation speed
- ~72 tok/s
- generation speed
- Cost per 1M output tokens
- $0.62
- per 1M output tokens
How Much VRAM Does Gemma 3 12B Pt Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 5.7 GB | 11.9 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~115 tok/s | |
| 5.9 GB | 12.1 GB | 3-bit small — 3-bit small quantization | Fits: 32K context · ~112 tok/s | |
| 6.5 GB | 12.7 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~100 tok/s | |
| 6.7 GB | 12.9 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 32K context · ~98 tok/s | |
| 8.0 GB | 14.2 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~82 tok/s | |
| 9.6 GB | 15.7 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~69 tok/s | |
| 11.1 GB | 17.3 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~59 tok/s | |
| 13.4 GB | 19.6 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~49 tok/s |
Selected Q4_K_M
4-bit medium quantization10.9GBat 16K context
- Weights
- 7.3 GB
- KV cache
- 3.3 GB
- Overhead
- 0.3 GB
Which Hardware Can Run Gemma 3 12B Pt?
Q4_K_M · 8.0 GBGemma 3 12B Pt (Q4_K_M) requires 8.0 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 11+ GB is recommended. 51 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080 Ti. 171 devices with unified memory can run Gemma 3 12B Pt, including Mac Studio M5 Ultra (256 GB), Mac Studio M5 Ultra (512 GB), Snapdragon X Plus (8-core) Copilot+ PC.
12 GB
12 GB tier →fits · 1.1 GB free · up to 16K ctx
16 GPUs
RTX 3080 Ti~74Simulate output on RTX 3080 TiRTX 3080 12GB~74Simulate output on RTX 3080 12GBRTX 5070~54Simulate output on RTX 507020 GB
fits · 9.1 GB free · up to 32K ctx
1 GPU
RX 7900 XT~60Simulate output on RX 7900 XT32–40 GB
fits · 21.1 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 9 device families
RTX 5090~145Simulate output on RTX 5090M5 Max*36 GB~40Simulate output on Mac Studio M5 Max (36 GB)M632 GB~15Simulate output on iMac 24" M6 (32 GB)RTX Spark N1X~24Simulate output on NVIDIA RTX Spark N1X Laptop (32 GB)Orin~17Simulate output on NVIDIA Jetson AGX Orin 32GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Gemma 3 12B Pt
Gemma 3 family · 20 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Mistral Nemo Base 2407Mistral AI | — | 11 GB | ~81 | same size |
| Baichuan2 13B Basebaichuan-inc | — | 11.5 GB | ~76 | same size |
| Llama 2 13B Chat HFMeta | — | 11.5 GB | ~76 | same size |
| Vicuna 13B V1.3LMSYS | — | 8.6 GB | ~76 | same size |
| NVIDIA Nemotron Nano 12B v2NVIDIA | — | 11 GB | ~81 | same size |
| Midm 2.0 Base InstructK-intelligence | — | 10.5 GB | ~86 | same size |
Fine-tunes & variants(2)
Specifications
- Publisher
- Family
- Gemma 3
- Parameters
- 12.2B
- Release Date
- 2025-03-01
- License
- Gemma Terms
Frequently Asked Questions
How much VRAM does Gemma 3 12B Pt need?
Gemma 3 12B Pt requires 8.0 GB of VRAM at Q4_K_M, or 26.8 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 12.2B × 4.8 bits ÷ 8 = 7.3 GB
KV Cache + Overhead ≈ 0.7 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Gemma 3 12B Pt?
Yes, at Q8_0 (13.4 GB) or lower. Higher quantizations like BF16 (26.8 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Gemma 3 12B Pt?
For Gemma 3 12B Pt, Q4_K_M (8.0 GB) offers the best balance of quality and VRAM usage. Q5_K_S (9.2 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 3.7 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Gemma 3 12B Pt on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 145 other Macs can run Gemma 3 12B Pt. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 8.0 GB of usable unified memory (RAM minus macOS overhead).
Can I run Gemma 3 12B Pt locally?
Yes — Gemma 3 12B Pt can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Gemma 3 12B Pt?
At Q4_K_M, Gemma 3 12B Pt can reach ~597 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~82 tok/s. 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 ÷ 8.04 × 0.65 = ~647 tok/s
Estimated speed at Q4_K_M (8.0 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Gemma 3 12B Pt?
At Q4_K_M, the download is about 7.31 GB. The full-precision BF16 version is 24.37 GB. The smallest option (IQ2_XXS) is 3.35 GB.
Which GPUs can run Gemma 3 12B Pt?
51 consumer GPUs can run Gemma 3 12B Pt at Q4_K_M. Top options include AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 6900 XT, AMD Radeon RX 6700 XT. 30 GPUs have plenty of headroom for comfortable inference.
Which devices can run Gemma 3 12B Pt?
171 devices with unified memory can run Gemma 3 12B Pt at Q4_K_M (8.0 GB), including AMD Ryzen 7 8845HS (Hawk Point) Laptop, AMD Ryzen AI 7 350 (Krackan Point) Laptop, AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.