spark s1 4b v6 needs about 6.1 GB of memory at Q8_0 (a 4.5 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 3 quants, from 3.5 GB (Q6_K) to 8.4 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Spark S1 4B v6 – a lightweight decision model for single‑token probability outputs
Spark S1 4B v6 is a Qwen3.5‑4B backbone model designed for local inference on NVIDIA DGX Spark. It returns a probability distribution over a fixed menu of options from the first token logits, without generating text. The model is released under Apache‑2.0 and is not a chat model.
Good at
- Fast single‑token decision making
- High accuracy on JevBench public items
- Low VRAM footprint with Q6_K quantization
Limits
- Only supports decision tasks, not general text classification
- No vision or multimodal support
- Requires calibrated temperature for accurate probabilities
Best for
- Automated support ticket routing
- Real‑time policy compliance checks
- Embedded decision engines in edge devices
Who it’s for. Developers building lightweight inference pipelines that need fast, single‑token probability outputs. The model is ideal for systems that require deterministic decision scores without full text generation.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q6_K | 3.5 GB | 5.0 GB | Fits | Get file → | |
| Q8_0 | 4.5 GB | 6.1 GB | Fits | Get file → | |
| BF16 | 8.4 GB | 10.0 GB | Too big | Get file → |
For 8 GB VRAM use Q6_K (3.46 GB file, ~5 GB VRAM at 8 k context). For 12 GB VRAM Q8_0 (4.48 GB file, ~6.1 GB VRAM) offers higher precision. BF16 requires 8.42 GB file and ~10 GB VRAM, suitable for larger GPUs.
What GPU do you need for spark s1 4b v6?
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: the best quant that fits is Q6_K (3.5 GB file, about 5.0 GB in use).
- 8 GB VRAM: the best quant that fits is Q8_0 (4.5 GB file, about 6.1 GB in use).
- 12 GB VRAM: the best quant that fits is BF16 (8.4 GB file, about 10.0 GB in use).
Which spark s1 4b v6 quant should you choose?
Q8_0 (4.5 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q6_K at 3.5 GB, with the quality trade-offs that come with fewer bits. Very close to the original quality for most uses, at a clearly larger size than 4-bit.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about spark s1 4b v6
How much VRAM does spark s1 4b v6 need?
About 6.1 GB at Q8_0 with an 8K context window. That is the 4.5 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 spark s1 4b v6 on an 8 GB GPU?
Yes, with Q8_0: the 4.5 GB file needs about 6.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of spark s1 4b v6 GGUF?
It depends on the quant: Q6_K is 3.5 GB, Q8_0 is 4.5 GB, BF16 is 8.4 GB.
What context length does spark s1 4b v6 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 spark s1 4b v6 GGUF?
The files are on Hugging Face in the abhishek085/spark-s1-4b-v6-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 spark s1 4b v6 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 Spark S1 4B v6 on 8 GB VRAM?
Yes, the Q6_K quantization uses about 5 GB VRAM at 8 k context, fitting within an 8 GB GPU.
Does the model support image inputs?
No, Spark S1 is a text‑only decision model and does not handle vision inputs.
Is the model compatible with Ollama?
Yes, Ollama 0.12+ supports the GGUF file and logprobs needed for decision extraction.