Dbrx Instruct needs about 81.7 GB of memory at Q4_K_M (a 79.9 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 22 quants, from 5.5 GB (Q3_K) to 139.8 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 5.4 GB.
Dbrx Instruct is a 208B-parameter Dbrx model in GGUF format, quantized by RichardErkhov. Its trained context length is 32K tokens. The repository has 22 quant files, from 5.5 GB (Q3_K) to 139.8 GB (Q8_0).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K | 5.5 GB | 7.4 GB | Tight | Get file → | |
| IQ3_M | 17.9 GB | 19.8 GB | Too big | Get file → | |
| Q3_K_S | 34.4 GB | 36.2 GB | Too big | Get file → | |
| Q2_K | 47.9 GB (2 parts) | 49.8 GB | Too big | Get file → | |
| IQ3_XS | 53.8 GB (2 parts) | 55.7 GB | Too big | Get file → | |
| IQ3_S | 56.8 GB (2 parts) | 58.6 GB | Too big | Get file → | |
| Q3_K_M | 63.1 GB (2 parts) | 65.0 GB | Too big | Get file → | |
| Q3_K_L | 68.4 GB (2 parts) | 70.2 GB | Too big | Get file → | |
| IQ4_XS | 70.9 GB (2 parts) | 72.8 GB | Too big | Get file → | |
| Q4_0 | 74.2 GB (2 parts) | 76.0 GB | Too big | Get file → | |
| IQ4_NL | 74.9 GB (2 parts) | 76.7 GB | Too big | Get file → | |
| Q4_K_S | 74.9 GB (2 parts) | 76.7 GB | Too big | Get file → | |
| Q4_K | 79.9 GB (3 parts) | 81.7 GB | Too big | Get file → | |
| Q4_K_M | 79.9 GB (3 parts) | 81.7 GB | Too big | Get file → | |
| Q4_1 | 82.4 GB (3 parts) | 84.2 GB | Too big | Get file → | |
| Q5_0 | 90.6 GB (3 parts) | 92.4 GB | Too big | Get file → | |
| Q5_K_S | 90.6 GB (3 parts) | 92.4 GB | Too big | Get file → | |
| Q5_K | 90.7 GB (3 parts) | 92.6 GB | Too big | Get file → | |
| Q5_K_M | 93.6 GB (3 parts) | 95.5 GB | Too big | Get file → | |
| Q5_1 | 96.8 GB (3 parts) | 98.6 GB | Too big | Get file → | |
| Q6_K | 108.0 GB (3 parts) | 109.8 GB | Too big | Get file → | |
| Q8_0 | 139.8 GB (4 parts) | 141.7 GB | Too big | Get file → |
What GPU do you need for Dbrx Instruct?
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: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: the best quant that fits is Q3_K (5.5 GB file, about 7.4 GB in use).
- 12 GB VRAM: the best quant that fits is Q3_K (5.5 GB file, about 7.4 GB in use).
- 16 GB VRAM: the best quant that fits is Q3_K (5.5 GB file, about 7.4 GB in use).
- 24 GB VRAM: the best quant that fits is IQ3_M (17.9 GB file, about 19.8 GB in use).
- 32 GB VRAM: the best quant that fits is IQ3_M (17.9 GB file, about 19.8 GB in use).
- 48 GB VRAM: the best quant that fits is Q3_K_S (34.4 GB file, about 36.2 GB in use).
Which Dbrx Instruct quant should you choose?
Q4_K_M is the usual starting point for Dbrx Instruct: it balances file size and quality well. Here it is a 79.9 GB file.
The Q8_0 file is 139.8 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q3_K at 5.5 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about Dbrx Instruct
How much VRAM does Dbrx Instruct need?
About 81.7 GB at Q4_K_M with an 8K context window. That is the 79.9 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 Dbrx Instruct on an 8 GB GPU?
Yes, with Q3_K: the 5.5 GB file needs about 7.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Dbrx Instruct GGUF?
It depends on the quant: Q3_K is 5.5 GB, IQ3_M is 17.9 GB, Q3_K_S is 34.4 GB, Q2_K is 47.9 GB, IQ3_XS is 53.8 GB, IQ3_S is 56.8 GB and 16 more.
What context length does Dbrx Instruct support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 1.3 GB at 8K and 5.4 GB at 32K.
Where can I download Dbrx Instruct GGUF?
The files are on Hugging Face in the RichardErkhov/databricks_-_dbrx-instruct-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 Dbrx Instruct 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.