FL

vllm-d3-flash

FlashGGUF by ggml-orgUpdated 2026-10-19
Download on Hugging Face
Parameters8.95B
Context256K
Layers32
Architectureqwen35
Smallest quant6.6 GB
Quick answer

vllm-d3-flash needs about 8.2 GB of memory at Q4_K_M (a 6.6 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 3 quants, from 6.6 GB (Q4_K_M) to 17.9 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

9B multimodal decision model for text, image, and video inputs

vllm-d3-flash is a 9B multimodal foundation model based on the Qwen‑35 architecture. It answers multiple-choice, yes/no, and scoring questions in a single call, returning probabilities for each option. The model supports text, images, and videos, and is optimized for fast inference on GPUs.

Good at

  • Fast multimodal decision making
  • Handles multiple questions per request
  • Supports image and video inputs

Limits

  • Requires 8‑k context for memory estimates
  • No text generation, only decision outputs
  • License is Apache‑2.0

Best for

  • Rapid decision support in applications
  • Batching multiple-choice queries
  • Integrating vision into decision pipelines

Who it’s for. Developers building decision‑oriented services that need quick, probabilistic answers from text, images, or videos. The model is ideal for systems that require a single API call to answer several related questions.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_K_M6.6 GB 8.2 GB
Too big Get file →
Q8_09.5 GB 11.1 GB
Too big Get file →
BF1617.9 GB 19.5 GB
Too big Get file →

For 8 GB VRAM, use the Q4_K_M quant (6.6 GB file, ~8.2 GB memory at 8k context). With 12 GB VRAM, Q8_0 (9.53 GB file, ~11.1 GB memory) is suitable. For larger budgets, BF16 (17.92 GB file, ~19.5 GB memory) offers higher precision.

What GPU do you need for vllm-d3-flash?

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: no quant fits at an 8K context.
  • 12 GB VRAM: the best quant that fits is Q8_0 (9.5 GB file, about 11.1 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (9.5 GB file, about 11.1 GB in use).
  • 24 GB VRAM: the best quant that fits is BF16 (17.9 GB file, about 19.5 GB in use).

Which vllm-d3-flash quant should you choose?

Q4_K_M is the usual starting point for vllm-d3-flash: it balances file size and quality well. Here it is a 6.6 GB file.

The Q8_0 file is 9.5 GB, about 1.4 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.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters8.95B
Layers32
Trained context262,144 tokens
KV cache (fp16)128 KB per token

Source

Quantized byggml-org
FamilyFlash
Licenseapache-2.0
Downloads (HF)17,213
Last updated2026-10-19

Frequently asked questions about vllm-d3-flash

How much VRAM does vllm-d3-flash need?

About 8.2 GB at Q4_K_M with an 8K context window. That is the 6.6 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 vllm-d3-flash on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 8.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of vllm-d3-flash GGUF?

It depends on the quant: Q4_K_M is 6.6 GB, Q8_0 is 9.5 GB, BF16 is 17.9 GB.

What context length does vllm-d3-flash 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 vllm-d3-flash GGUF?

The files are on Hugging Face in the ggml-org/vllm-d3-flash-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 vllm-d3-flash 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 vllm-d3-flash on 8 GB VRAM?

Yes, the Q4_K_M quant file is 6.6 GB and requires about 8.2 GB of VRAM at 8k context.

Does the model support text generation?

No, it only returns decision probabilities for choice, yes/no, and score questions.

What file sizes are available for different quantizations?

Q4_K_M is 6.6 GB, Q8_0 is 9.53 GB, and BF16 is 17.92 GB.

Other GGUFs of d3-flash

Similar models

Browse all →

More from ggml-org

All by ggml-org →

More Flash models

All Flash →

Other small models

All small models →