Nemotron 3.5 Embed 8B BF16 needs about 6.5 GB of memory at Q4_K_M (a 4.9 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 35 quants, from 0.0 GB to 31.8 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 4.6 GB.
Nemotron 3.5 Embed 8B BF16 – multimodal embedding model
Nemotron 3.5 Embed 8B BF16 is a multimodal embedding model that accepts text or document images and outputs 4096‑dimensional vectors. It supports up to 262144 context tokens and is trained on a wide range of languages for retrieval tasks.
Good at
- Multilingual document retrieval
- Image‑to‑text embedding
- Dense vector search
Limits
- No vision support in current llama.cpp embedding endpoint
- Large memory requirement for high‑quality quants
- License governed by OpenMDW 1.1
Best for
- Building search engines over mixed text and image corpora
- Creating embeddings for vector databases
- Retrieval‑augmented generation pipelines
Who it’s for. Developers and researchers needing a dense embedding model for multimodal retrieval. The model is suitable for commercial or non‑commercial use under the OpenMDW 1.1 license.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 1.7 GB | Fits | Get file → | |
| Q1_0 | 1.5 GB | 3.1 GB | Fits | Get file → | |
| IQ1_S | 1.9 GB | 3.6 GB | Fits | Get file → | |
| TQ1_0 | 2.0 GB | 3.7 GB | Fits | Get file → | |
| IQ1_M | 2.1 GB | 3.7 GB | Fits | Get file → | |
| IQ2_XXS | 2.3 GB | 4.0 GB | Fits | Get file → | |
| TQ2_0 | 2.4 GB | 4.0 GB | Fits | Get file → | |
| Q2_0 | 2.5 GB | 4.2 GB | Fits | Get file → | |
| IQ2_XS | 2.6 GB | 4.2 GB | Fits | Get file → | |
| IQ2_S | 2.7 GB | 4.3 GB | Fits | Get file → | |
| IQ2_M | 2.9 GB | 4.5 GB | Fits | Get file → | |
| Q2_K_S | 3.0 GB | 4.6 GB | Fits | Get file → | |
| Q2_K | 3.2 GB | 4.8 GB | Fits | Get file → | |
| IQ3_XXS | 3.2 GB | 4.9 GB | Fits | Get file → | |
| IQ3_XS | 3.5 GB | 5.1 GB | Fits | Get file → | |
| Q3_K_S | 3.6 GB | 5.3 GB | Fits | Get file → | |
| IQ3_S | 3.7 GB | 5.3 GB | Fits | Get file → | |
| IQ3_M | 3.8 GB | 5.4 GB | Fits | Get file → | |
| Q3_K_M | 4.0 GB | 5.7 GB | Fits | Get file → | |
| Q3_K_L | 4.3 GB | 6.0 GB | Fits | Get file → | |
| IQ4_XS | 4.4 GB | 6.1 GB | Fits | Get file → | |
| Q4_0 | 4.6 GB | 6.3 GB | Fits | Get file → | |
| Q4_K_S | 4.7 GB | 6.3 GB | Fits | Get file → | |
| IQ4_NL | 4.7 GB | 6.3 GB | Fits | Get file → | |
| Q4_K_M | 4.9 GB | 6.5 GB | Fits | Get file → | |
| Q4_1 | 5.1 GB | 6.7 GB | Fits | Get file → | |
| Q5_0 | 5.5 GB | 7.2 GB | Fits | Get file → | |
| Q5_K_S | 5.5 GB | 7.2 GB | Fits | Get file → | |
| Q5_K_M | 5.7 GB | 7.3 GB | Tight | Get file → | |
| Q5_1 | 6.0 GB | 7.7 GB | Tight | Get file → | |
| Q6_K | 6.5 GB | 8.2 GB | Too big | Get file → | |
| Q8_0 | 8.5 GB | 10.1 GB | Too big | Get file → | |
| BF16 | 15.9 GB | 17.6 GB | Too big | Get file → | |
| F16 | 15.9 GB | 17.6 GB | Too big | Get file → | |
| F32 | 31.8 GB | 33.5 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_0 (file size 4.62 GB, memory 6.3 GB). For 4 GB VRAM, use Q3_K_S (file size 3.64 GB, memory 5.3 GB). For 2 GB VRAM, use Q2_0 (file size 2.54 GB, memory 4.2 GB).
What GPU do you need for Nemotron 3.5 Embed 8B BF16?
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 TQ2_0 (2.4 GB file, about 4.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q3_K_L (4.3 GB file, about 6.0 GB in use).
- 8 GB VRAM: the best quant that fits is Q5_1 (6.0 GB file, about 7.7 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (8.5 GB file, about 10.1 GB in use).
- 24 GB VRAM: the best quant that fits is BF16 (15.9 GB file, about 17.6 GB in use).
- 32 GB VRAM: the best quant that fits is BF16 (15.9 GB file, about 17.6 GB in use).
- 48 GB VRAM: the best quant that fits is F32 (31.8 GB file, about 33.5 GB in use).
Which Nemotron 3.5 Embed 8B BF16 quant should you choose?
Q4_K_M is the usual starting point for Nemotron 3.5 Embed 8B BF16: it balances file size and quality well. Here it is a 4.9 GB file.
The Q8_0 file is 8.5 GB, about 1.7 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 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 Nemotron 3.5 Embed 8B BF16
How much VRAM does Nemotron 3.5 Embed 8B BF16 need?
About 6.5 GB at Q4_K_M with an 8K context window. That is the 4.9 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 Nemotron 3.5 Embed 8B BF16 on an 8 GB GPU?
Yes, with Q5_1: the 6.0 GB file needs about 7.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Nemotron 3.5 Embed 8B BF16 GGUF?
It depends on the quant: file is 0.0 GB, Q1_0 is 1.5 GB, IQ1_S is 1.9 GB, TQ1_0 is 2.0 GB, IQ1_M is 2.1 GB, IQ2_XXS is 2.3 GB and 29 more.
What context length does Nemotron 3.5 Embed 8B BF16 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 4.6 GB at 32K.
Where can I download Nemotron 3.5 Embed 8B BF16 GGUF?
The files are on Hugging Face in the W4D/Nemotron-3.5-Embed-8B-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 Nemotron 3.5 Embed 8B BF16 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 this model on 8 GB VRAM?
Yes, the Q4_0 quant uses approximately 6.3 GB of VRAM at 8K context, which fits within an 8 GB budget.
Does the model support image inputs in llama.cpp?
Currently, image inputs are not supported on the embedding endpoint in llama.cpp; text-only usage is available.
What license applies to the GGUF conversion?
The GGUF conversion is governed by the OpenMDW 1.1 license, same as the original model.