Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized — Hardware Requirements & GPU Compatibility
Vision
Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized is a 12.2B-parameter open language model from ibyteohdear in the Gemma 3 family. It supports a context window of up to 131,072 tokens. At Q4_K_M it needs about 8.37 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_0 · 16K)Best on 24 GB (Q4_K_M · ~78 tok/s · 128K)Macs from 16 GB (Q4_K_M · ~17 tok/s)
Gemma Terms
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized
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/Lightricks-gemma-3-12b-it-qat-q4_0-unquantized-i1-GGUFQ4_K_M fits fully on the RTX 4090 · 14.8 GB free · 16K context
Download: Lightricks-gemma-3-12b-it-qat-q4_0-unquantized.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 3080 Ti (12 GB) (2 GB headroom) · Q4_K_M
- Generation speed
- ~71 tok/s
- generation speed
- Cost per 1M output tokens
- $0.48
- per 1M output tokens
How Much VRAM Does Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 6.2 GB | 8.3 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 128K context · ~105 tok/s | |
| 7 GB | 9.1 GB | 3-bit medium — 3-bit medium quantization | Fits: 128K context · ~94 tok/s | |
| 7.2 GB | 9.2 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 128K context · ~92 tok/s | |
| 8.4 GB | 10.4 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 128K context · ~78 tok/s | |
| 9.7 GB | 11.8 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 128K context · ~67 tok/s | |
| 11.1 GB | 13.2 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 128K context · ~59 tok/s | |
| 13.2 GB | 15.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 128K context · ~50 tok/s | |
| 25.4 GB | 27.5 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 26.2 GB |
Selected Q4_K_M
4-bit medium quantization9.2GBat 16K context
- Weights
- 7.3 GB
- KV cache
- 1.6 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 14.8 GB free · ~78 tok/s
≥50 tok/s needs ≥643 GB/s · RTX 5090 ~139, RTX 3090 Ti ~78
Which Hardware Can Run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
Q4_K_M · 8.4 GBLightricks Gemma 3 12B IT Qat Q4 0 Unquantized (Q4_K_M) requires 8.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 11+ GB is recommended. Using the full 131K context window can add up to 9.6 GB, bringing total usage to 18.0 GB. 40 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080 Ti. 150 devices with unified memory can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M2 Pro (16 GB).
12 GB
12 GB tier →fits · 2.8 GB free · up to 32K ctx
10 GPUs
RTX 3080 Ti~71Simulate output on RTX 3080 TiRTX 5070~52Simulate output on RTX 5070RTX 4070~39Simulate output on RTX 407020 GB
fits · 10.8 GB free · up to 128K ctx
1 GPU
RX 7900 XT~57Simulate output on RX 7900 XT32–40 GB
fits · 22.8 GB free · up to 128K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~139Simulate output on RTX 5090M5 Max36 GB~39Simulate output on Mac Studio M5 Max (36 GB)M532 GB~13Simulate output on MacBook Air 13" M5 (32 GB)Orin~16Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~11Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized
Gemma 3 family · 20 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Gemma 4 12BGoogle | — | 9 GB | ~80 | same size |
| 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 |
Specifications
- Publisher
- ibyteohdear
- Family
- Gemma 3
- Parameters
- 12.2B
- Architecture
- Gemma3ForConditionalGeneration
- Context Length
- 131,072 tokens
- Vocabulary Size
- 262,208
- Release Date
- 2026-05-27
- License
- Gemma Terms
Frequently Asked Questions
How much VRAM does Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized need?
Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized requires 8.4 GB of VRAM at Q4_K_M, or 25.4 GB at BF16. Full 131K context adds up to 9.6 GB (18.0 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 12.2B × 4.8 bits ÷ 8 = 7.3 GB
KV Cache + Overhead ≈ 1.1 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.
KV Cache + Overhead ≈ 10.7 GB (at full 131K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
Yes, at Q8_0 (13.2 GB) or lower. Higher quantizations like BF16 (25.4 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
For Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized, Q4_K_M (8.4 GB) offers the best balance of quality and VRAM usage. Q5_K_M (9.7 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 6.2 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 8.4 GB of usable unified memory (RAM minus macOS overhead).
Can I run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized locally?
Yes — Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
At Q4_K_M, Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized can reach ~574 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~78 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 ÷ 8.37 × 0.65 = ~621 tok/s
Estimated speed at Q4_K_M (8.4 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
At Q4_K_M, the download is about 7.31 GB. The full-precision BF16 version is 24.37 GB. The smallest option (Q2_K) is 5.18 GB.
Which GPUs can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
40 consumer GPUs can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized 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. 26 GPUs have plenty of headroom for comfortable inference.
Which devices can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized?
150 devices with unified memory can run Lightricks Gemma 3 12B IT Qat Q4 0 Unquantized at Q4_K_M (8.4 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (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.