DeepSeek GRM 16B needs about 11.3 GB of memory at Q4_K_M (a 10.4 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 25 quants, from 0.0 GB to 14.1 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.
DeepSeek GRM 16B – a 15.7B parameter generative reward model
DeepSeek GRM 16B is a generative reward model based on DeepSeek-V2-Lite. It uses Self‑Principled Critique Tuning to produce principle‑driven critiques and scores. The model supports tool usage and is available in multiple GGUF quantizations.
Good at
- Tool‑enabled reward generation
- Principle‑driven critique
- Score emission
Limits
- No vision support
- Requires 163840 context tokens
- License is other
Best for
- Local reward model deployment
- Offline data collection
- Online RL for LLMs
Who it’s for. Developers and researchers needing a local generative reward model with tool support. The model is suitable for those who prefer GGUF format and can handle the 16B parameter size.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 1.0 GB | Fits | Get file → | |
| IQ1_S | 5.0 GB | 6.0 GB | Fits | Get file → | |
| IQ1_M | 5.2 GB | 6.2 GB | Fits | Get file → | |
| IQ2_XXS | 5.6 GB | 6.6 GB | Fits | Get file → | |
| IQ2_XS | 6.0 GB | 7.0 GB | Fits | Get file → | |
| IQ2_S | 6.0 GB | 7.0 GB | Fits | Get file → | |
| IQ2_M | 6.3 GB | 7.3 GB | Tight | Get file → | |
| Q2_K | 6.4 GB | 7.4 GB | Tight | Get file → | |
| Q2_K_S | 6.5 GB | 7.4 GB | Tight | Get file → | |
| IQ3_XXS | 7.0 GB | 8.0 GB | Tight | Get file → | |
| IQ3_XS | 7.1 GB | 8.1 GB | Too big | Get file → | |
| IQ3_S | 7.5 GB | 8.5 GB | Too big | Get file → | |
| Q3_K_S | 7.5 GB | 8.5 GB | Too big | Get file → | |
| IQ3_M | 7.6 GB | 8.5 GB | Too big | Get file → | |
| Q3_K_M | 8.1 GB | 9.1 GB | Too big | Get file → | |
| Q3_K_L | 8.5 GB | 9.4 GB | Too big | Get file → | |
| IQ4_XS | 8.6 GB | 9.6 GB | Too big | Get file → | |
| IQ4_NL | 8.9 GB | 9.9 GB | Too big | Get file → | |
| Q4_0 | 8.9 GB | 9.9 GB | Too big | Get file → | |
| Q4_K_S | 9.5 GB | 10.5 GB | Too big | Get file → | |
| Q4_1 | 9.9 GB | 10.9 GB | Too big | Get file → | |
| Q4_K_M | 10.4 GB | 11.3 GB | Too big | Get file → | |
| Q5_K_S | 11.1 GB | 12.1 GB | Too big | Get file → | |
| Q5_K_M | 11.9 GB | 12.8 GB | Too big | Get file → | |
| Q6_K | 14.1 GB | 15.1 GB | Too big | Get file → |
For 8 GB VRAM, the Q4_K_M quant (~10.37 GB file, 11.3 GB memory at 8k context) is the largest that fits. If 12 GB VRAM is available, Q5_K_M (~11.85 GB file, 12.8 GB memory) can be used. Smaller quants such as IQ1_S or IQ2_XXS provide lower memory usage at the cost of precision.
What GPU do you need for DeepSeek GRM 16B?
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 the main file (0.0 GB file, about 1.0 GB in use).
- 6 GB VRAM: the best quant that fits is IQ1_S (5.0 GB file, about 6.0 GB in use).
- 8 GB VRAM: the best quant that fits is IQ3_XXS (7.0 GB file, about 8.0 GB in use).
- 12 GB VRAM: the best quant that fits is Q4_K_M (10.4 GB file, about 11.3 GB in use).
- 16 GB VRAM: the best quant that fits is Q6_K (14.1 GB file, about 15.1 GB in use).
Which DeepSeek GRM 16B quant should you choose?
Q4_K_M is the usual starting point for DeepSeek GRM 16B: it balances file size and quality well. Here it is a 10.4 GB file.
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about DeepSeek GRM 16B
How much VRAM does DeepSeek GRM 16B need?
About 11.3 GB at Q4_K_M with an 8K context window. That is the 10.4 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 DeepSeek GRM 16B on an 8 GB GPU?
Yes, with IQ3_XXS: the 7.0 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of DeepSeek GRM 16B GGUF?
It depends on the quant: file is 0.0 GB, IQ1_S is 5.0 GB, IQ1_M is 5.2 GB, IQ2_XXS is 5.6 GB, IQ2_XS is 6.0 GB, IQ2_S is 6.0 GB and 19 more.
What context length does DeepSeek GRM 16B support?
The GGUF header lists a trained context length of 163,840 tokens (160K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.
Where can I download DeepSeek GRM 16B GGUF?
The files are on Hugging Face in the mradermacher/DeepSeek-GRM-16B-i1-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 DeepSeek GRM 16B use?
Hugging Face lists the license as other. 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 DeepSeek GRM 16B on 8 GB VRAM?
The Q4_K_M quant uses 11.3 GB memory at 8k context, so it exceeds 8 GB VRAM. Smaller quants like IQ1_S use 6 GB memory and fit within 8 GB VRAM.
What context length does the model support?
The trained context length is 163840 tokens.
Does the model support vision or audio?
No, the model has no vision or audio capabilities.