Run SulphurAI/Sulphur-2-base locally

⬇ 799,631 ❤ 1808
Parameters9.2B
Context262,144

SulphurAI/Sulphur-2-base is a large language model with 9.2 billion parameters, built on the qwen35 architecture. It has been downloaded 799,631 times.

To run SulphurAI/Sulphur-2-base locally at a 4,096-token context, its quantized versions need between 2.13 GB (GGUF, lowest quality) and 19.26 GB (Q8_0, highest quality) of memory, weights plus KV cache and a system margin included.

For most users the best balance is GGUF, needing about 2.13 GB. That means SulphurAI/Sulphur-2-base fits entirely in the VRAM of a 6 GB GPU or larger, running fully on the GPU.

→ Guide: How much VRAM do you need?

All quantizations

Quant.Bits QualityWeights KVTotal Speed~Verdict
GGUF 0.8 Very low 0.86 GB 0.47 GB 2.13 GB 466.0 t/s Fits in VRAM
BF16 0.8 Very low 0.86 GB 0.47 GB 2.13 GB 466.0 t/s Fits in VRAM
Q8_0 16.8 Excellent 17.99 GB 0.47 GB 19.26 GB 2.8 t/s Offload

KV cache computed from the model's exact architecture. Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

How much VRAM do you need to run SulphurAI/Sulphur-2-base?

You need about 2.13 GB of VRAM to run SulphurAI/Sulphur-2-base entirely on the GPU using the GGUF quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run SulphurAI/Sulphur-2-base on an 8 GB GPU?

Yes. With 8 GB of VRAM you can run SulphurAI/Sulphur-2-base fully on the GPU using GGUF (about 2.13 GB).

Can I run SulphurAI/Sulphur-2-base on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run SulphurAI/Sulphur-2-base fully on the GPU using GGUF (about 2.13 GB).

Can I run SulphurAI/Sulphur-2-base on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run SulphurAI/Sulphur-2-base fully on the GPU using Q8_0 (about 19.26 GB).

What is the best quantization for SulphurAI/Sulphur-2-base?

If memory allows, higher bits-per-weight means better quality. A common sweet spot is a Q4_K_M or Q5_K_M quantization, which keeps most of the quality while roughly halving the memory versus 8-bit. Pick the highest quantization that still fits in your VRAM.