FA

Falcon

FalconGGUF by mradermacherUpdated 2024-06-02
Download on Hugging Face
Parameters5.47B
Context2K
Layers28
Architecturefalcon
Smallest quant2.2 GB
Quick answer

Falcon needs about 4.9 GB of memory at Q4_K_M (a 3.4 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 15 quants, from 2.2 GB (Q2_K) to 10.9 GB (F16).

Falcon2‑5.5B Dutch GGUF model

Falcon2‑5.5B Dutch is a 5.47B parameter language model fine‑tuned on Dutch data. It is available in multiple GGUF quantizations for efficient inference. The model supports 2048 token context and is released under Apache‑2.0.

Good at

  • Dutch language tasks
  • Text generation
  • Summarization

Limits

  • No vision support
  • Limited to 2048 token context
  • Requires PyTorch 2.0 for transformers

Best for

  • Chatbot prototypes
  • Research on Dutch NLP
  • Fine‑tuning for Dutch applications

Who it’s for. Researchers and developers needing a Dutch‑capable LLM for experimentation or prototyping. The model is suitable for academic projects and low‑resource fine‑tuning.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K2.2 GB 2.9 GB
Fits Get file →
IQ3_XS2.5 GB 3.2 GB
Fits Get file →
IQ3_S2.5 GB 3.3 GB
Fits Get file →
Q3_K_S2.5 GB 3.3 GB
Fits Get file →
IQ3_M2.6 GB 3.4 GB
Fits Get file →
Q3_K_M2.7 GB 3.5 GB
Fits Get file →
Q3_K_L2.9 GB 3.7 GB
Fits Get file →
IQ4_XS3.1 GB 3.8 GB
Fits Get file →
Q4_K_S3.2 GB 3.9 GB
Fits Get file →
Q4_K_M3.4 GB 4.2 GB
Fits Get file →
Q5_K_S3.9 GB 4.6 GB
Fits Get file →
Q5_K_M4.1 GB 4.8 GB
Fits Get file →
Q6_K4.6 GB 5.3 GB
Fits Get file →
Q8_05.8 GB 6.5 GB
Fits Get file →
F1610.9 GB 11.7 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (file size 3.43 GB, memory 4.9 GB). For 12 GB VRAM, Q5_K_M (file size 4.08 GB, memory 5.5 GB) is a good balance. For 16 GB VRAM, Q6_K (file size 4.55 GB, memory 6 GB) offers higher quality.

What GPU do you need for Falcon?

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 IQ3_S (2.5 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is Q6_K (4.6 GB file, about 6.0 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (5.8 GB file, about 7.2 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (5.8 GB file, about 7.2 GB in use).
  • 16 GB VRAM: the best quant that fits is F16 (10.9 GB file, about 12.4 GB in use).

Which Falcon quant should you choose?

Q4_K_M is the usual starting point for Falcon: it balances file size and quality well. Here it is a 3.4 GB file.

The Q8_0 file is 5.8 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 Q2_K at 2.2 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturefalcon
Parameters5.47B
Layers28
Trained context2,048 tokens
KV cache (fp16)112 KB per token

Source

Quantized bymradermacher
FamilyFalcon
Licenseapache-2.0
Downloads (HF)865
Last updated2024-06-02

Frequently asked questions about Falcon

How much VRAM does Falcon need?

About 4.9 GB at Q4_K_M with an 8K context window. That is the 3.4 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 Falcon on an 8 GB GPU?

Yes, with Q8_0: the 5.8 GB file needs about 7.2 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Falcon GGUF?

It depends on the quant: Q2_K is 2.2 GB, IQ3_XS is 2.5 GB, IQ3_S is 2.5 GB, Q3_K_S is 2.5 GB, IQ3_M is 2.6 GB, Q3_K_M is 2.7 GB and 9 more.

What context length does Falcon support?

The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 0.9 GB at 8K and 0.2 GB at 2K.

Where can I download Falcon GGUF?

The files are on Hugging Face in the mradermacher/Falcon2-5.5B-Dutch-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 Falcon 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 Falcon2‑5.5B Dutch on 8 GB VRAM?

Yes, the Q4_K_M quant uses about 4.9 GB of VRAM at 8 k context, which fits within 8 GB.

Does the model support other languages besides Dutch?

It was trained on Dutch and several other European languages, but performance is best in Dutch.

Is the model suitable for production use?

The model is intended for research and prototyping; production use requires careful risk assessment.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Falcon models

All Falcon →

Other small models

All small models →