Clef needs about 22.2 GB of memory at Q4_K_M (a 19.5 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 3 quants, from 19.5 GB (Q4_K_M) to 54.1 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 8.6 GB.
27B multimodal decision model for structured queries
Clef is a 27B multimodal model that processes text, JSON, images, or video to produce probabilities for each option in a set of questions. It outputs one logit per option, with no free‑form text generation. The model is built on Qwen/Qwen3.8‑27B and includes a vision encoder.
Good at
- Structured decision making
- Multimodal input handling
- Fast inference with single forward pass
Limits
- Requires 8 k context for memory estimates
- No free‑form text generation
- Vision support needs pillow
Best for
- Automated decision systems
- Batch evaluation of structured queries
- Integrating with Jev/SystemOne APIs
Who it’s for. Developers building decision‑making services that need reliable probability outputs for multiple choice, true/false, or score questions. Suitable for applications that combine text and visual data without requiring text generation.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 19.5 GB | 22.2 GB | Too big | Get file → | |
| Q8_0 | 28.7 GB | 31.4 GB | Too big | Get file → | |
| BF16 | 54.1 GB | 56.7 GB | Too big | Get file → |
For 8 GB VRAM use Q4_K_M (19.23 GB file, ~21.9 GB memory at 8 k context). For 12 GB VRAM use Q8_0 (28.73 GB file, ~31.4 GB memory). BF16 requires 54.06 GB file and ~56.7 GB memory, suitable for large‑memory setups.
What GPU do you need for Clef?
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: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: the best quant that fits is Q4_K_M (19.5 GB file, about 22.2 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (28.7 GB file, about 31.4 GB in use).
- 48 GB VRAM: the best quant that fits is Q8_0 (28.7 GB file, about 31.4 GB in use).
Which Clef quant should you choose?
Q4_K_M is the usual starting point for Clef: it balances file size and quality well. Here it is a 19.5 GB file.
The Q8_0 file is 28.7 GB, about 1.5 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
Source
Frequently asked questions about Clef
How much VRAM does Clef need?
About 22.2 GB at Q4_K_M with an 8K context window. That is the 19.5 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 Clef on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 22.2 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Clef GGUF?
It depends on the quant: Q4_K_M is 19.5 GB, Q8_0 is 28.7 GB, BF16 is 54.1 GB.
What context length does Clef support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.6 GB at 32K.
Where can I download Clef GGUF?
The files are on Hugging Face in the ggml-org/Clef-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 Clef 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 Clef on 8 GB VRAM?
Use the Q4_K_M quantized file; it requires about 21.9 GB of VRAM at 8 k context, so 8 GB is insufficient.
Does Clef support image input?
Yes, image and video inputs are supported via pillow and the media_kwargs processor.
What license applies to Clef?
Clef is released under the Apache‑2.0 license.