gemma 4 12b it MTP needs about 1.9 GB of memory at Q4_K_M (a 0.3 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 16 quants, from 0.3 GB (Q1_0) to 0.4 GB (Q6_K_M). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Gemma 4 12B IT MTP – a lightweight, multi‑token draft model for fast inference
Gemma 4 12B IT MTP is a quantized draft model that extends the base Gemma 4 12B instruction‑tuned assistant. It supports a context window of 262144 tokens and is designed for speculative decoding pipelines. The model is released under Apache 2.0.
Good at
- Fast speculative decoding
- Low‑latency local inference
- Compatible with llama.cpp
Limits
- No vision support in this draft model
- Requires a base model to function
- Large context window may increase VRAM usage
Best for
- On‑device chat assistants
- Low‑latency code generation
- Edge‑device inference
Who it’s for. Developers building low‑latency, on‑device LLM applications that can pair a draft model with a full base model. The MTP design is ideal for mobile and laptop deployments where VRAM is limited.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q1_0 | 0.3 GB | 1.8 GB | Fits | Get file → | |
| Q2_K_M | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_S | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_M | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_L | 0.3 GB | 1.9 GB | Fits | Get file → | |
| IQ4_XS | 0.3 GB | 1.9 GB | Fits | Get file → | |
| IQ4_NL | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q4_0 | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q4_K_S | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q4_K_M | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q4_1 | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q5_0 | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q5_K_S | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q5_K_M | 0.3 GB | 1.9 GB | Fits | Get file → | |
| Q5_1 | 0.4 GB | 1.9 GB | Fits | Get file → | |
| Q6_K_M | 0.4 GB | 1.9 GB | Fits | Get file → |
For 8 GB VRAM, use Q1_0 (0.26 GB file, ~1.8 GB memory at 8 k context). For 12 GB VRAM, Q4_K_M (0.33 GB file, ~1.9 GB memory). Higher quants like Q5_K_M or Q6_K_M increase file size to 0.35–0.36 GB and memory to ~1.9 GB, suitable for 16 GB VRAM setups.
What GPU do you need for gemma 4 12b it MTP?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q6_K_M (0.4 GB file, about 1.9 GB in use).
Which gemma 4 12b it MTP quant should you choose?
Q4_K_M is the usual starting point for gemma 4 12b it MTP: it balances file size and quality well. Here it is a 0.3 GB file.
If memory is tight, the smallest file is Q1_0 at 0.3 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about gemma 4 12b it MTP
How much VRAM does gemma 4 12b it MTP need?
About 1.9 GB at Q4_K_M with an 8K context window. That is the 0.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run gemma 4 12b it MTP on an 8 GB GPU?
Yes, with Q6_K_M: the 0.4 GB file needs about 1.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of gemma 4 12b it MTP GGUF?
It depends on the quant: Q1_0 is 0.3 GB, Q2_K_M is 0.3 GB, Q3_K_S is 0.3 GB, Q3_K_M is 0.3 GB, Q3_K_L is 0.3 GB, IQ4_XS is 0.3 GB and 10 more.
What context length does gemma 4 12b it MTP support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download gemma 4 12b it MTP GGUF?
The files are on Hugging Face in the cloudnathan5/gemma-4-12b-it-MTP-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does gemma 4 12b it MTP use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run this model on 8 GB VRAM?
Yes, the Q1_0 quant requires about 1.8 GB memory at 8 k context, fitting within 8 GB VRAM.
What context length does this model support?
It supports up to 262144 tokens, the maximum context window for Gemma 4 12B.
Is this model vision‑capable?
No, the MTP draft model handles only text; vision requires the base model.