SK

SkyCaptioner-V1

SkycaptionerGGUF by MungertUpdated 2025-09-24
Download on Hugging Face
Parameters7.62B
Context125K
Layers28
Architectureqwen2vl
Smallest quant2.1 GB
Quick answer

SkyCaptioner-V1 needs about 5.7 GB of memory at Q4_K_M (a 4.8 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 31 quants, from 2.1 GB (TQ1_0) to 15.2 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

SkyCaptioner‑V1: 7.62B Qwen2VL model for image captioning

SkyCaptioner‑V1 is a 7.62 billion‑parameter Qwen2VL model optimized for image captioning. It supports tool usage and is released under the Apache‑2.0 license. The model was quantized with llama.cpp and offers a range of quantization options.

Good at

  • Image captioning
  • Tool integration
  • Low‑bit quantization

Limits

  • No vision support beyond image captioning
  • Requires 8 k context
  • License is Apache‑2.0

Best for

  • Captioning images in local inference
  • Generating tool‑enabled prompts
  • Running on modest GPUs

Who it’s for. Developers and researchers needing a lightweight image captioning model that can run locally with llama.cpp. The model is suitable for projects that require tool usage and can operate within the provided memory constraints.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
TQ1_02.1 GB 3.1 GB
Fits Get file →
IQ1_S2.4 GB 3.4 GB
Fits Get file →
TQ2_02.4 GB 3.4 GB
Fits Get file →
IQ1_M2.6 GB 3.5 GB
Fits Get file →
IQ2_XXS2.7 GB 3.6 GB
Fits Get file →
IQ2_XS2.8 GB 3.8 GB
Fits Get file →
IQ2_S2.9 GB 3.9 GB
Fits Get file →
Q2_K_S2.9 GB 3.9 GB
Fits Get file →
IQ2_M3.0 GB 4.0 GB
Fits Get file →
Q2_K_M3.3 GB 4.2 GB
Fits Get file →
IQ3_XXS3.3 GB 4.2 GB
Fits Get file →
IQ3_XS3.4 GB 4.4 GB
Fits Get file →
IQ3_S3.6 GB 4.5 GB
Fits Get file →
IQ3_M3.6 GB 4.6 GB
Fits Get file →
Q3_K_S3.6 GB 4.6 GB
Fits Get file →
Q3_K_M4.0 GB 5.0 GB
Fits Get file →
IQ4_XS4.2 GB 5.2 GB
Fits Get file →
Q4_04.3 GB 5.3 GB
Fits Get file →
IQ4_NL4.4 GB 5.4 GB
Fits Get file →
Q4_K_S4.6 GB 5.6 GB
Fits Get file →
Q4_14.8 GB 5.7 GB
Fits Get file →
Q4_K_M4.8 GB 5.7 GB
Fits Get file →
Q5_05.2 GB 6.2 GB
Fits Get file →
Q5_K_S5.5 GB 6.4 GB
Fits Get file →
Q5_K_M5.5 GB 6.5 GB
Fits Get file →
Q5_15.7 GB 6.7 GB
Fits Get file →
Q6_K_M6.3 GB 7.2 GB
Tight Get file →
Q8_08.1 GB 9.1 GB
Too big Get file →
BF1611.3 GB 12.3 GB
Too big Get file →
F1611.3 GB 12.3 GB
Too big Get file →
BF1615.2 GB 16.2 GB
Too big Get file →

For 8 GB VRAM use the TQ1_0 or IQ1_S quantizations, each about 2.14 GB or 2.41 GB. If 12 GB VRAM is available, IQ2_M or Q2_K_M (~3.04 GB or 3.26 GB) provide higher precision. For 16 GB VRAM, Q3_K_M or IQ4_XS (~4 GB or 4.22 GB) are recommended.

What GPU do you need for SkyCaptioner-V1?

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 IQ2_M (3.0 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is Q4_K_M (4.8 GB file, about 5.7 GB in use).
  • 8 GB VRAM: the best quant that fits is Q6_K_M (6.3 GB file, about 7.2 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (8.1 GB file, about 9.1 GB in use).
  • 16 GB VRAM: the best quant that fits is BF16 (11.3 GB file, about 12.3 GB in use).
  • 24 GB VRAM: the best quant that fits is BF16 (15.2 GB file, about 16.2 GB in use).

Which SkyCaptioner-V1 quant should you choose?

Q4_K_M is the usual starting point for SkyCaptioner-V1: it balances file size and quality well. Here it is a 4.8 GB file.

The Q8_0 file is 8.1 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 TQ1_0 at 2.1 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen2vl
Parameters7.62B
Layers28
Trained context128,000 tokens
KV cache (fp16)56 KB per token

Source

Quantized byMungert
Licenseapache-2.0
Downloads (HF)848
Last updated2025-09-24

Frequently asked questions about SkyCaptioner-V1

How much VRAM does SkyCaptioner-V1 need?

About 5.7 GB at Q4_K_M with an 8K context window. That is the 4.8 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 SkyCaptioner-V1 on an 8 GB GPU?

Yes, with Q6_K_M: the 6.3 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 SkyCaptioner-V1 GGUF?

It depends on the quant: TQ1_0 is 2.1 GB, IQ1_S is 2.4 GB, TQ2_0 is 2.4 GB, IQ1_M is 2.6 GB, IQ2_XXS is 2.7 GB, IQ2_XS is 2.8 GB and 25 more.

What context length does SkyCaptioner-V1 support?

The GGUF header lists a trained context length of 128,000 tokens (125K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.

Where can I download SkyCaptioner-V1 GGUF?

The files are on Hugging Face in the Mungert/SkyCaptioner-V1-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 SkyCaptioner-V1 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 SkyCaptioner‑V1 on 8 GB VRAM?

Yes, the TQ1_0 quantization is 2.14 GB and requires about 3.1 GB memory at 8 k context, fitting comfortably on 8 GB VRAM.

Does the model support vision beyond captioning?

No, the model is designed specifically for image captioning and does not support other vision tasks.

Is the model free to use?

Yes, it is released under the Apache‑2.0 license, allowing free use and modification.

Similar models

Browse all →

More from Mungert

All by Mungert →

More Skycaptioner models

All Skycaptioner →

Other small models

All small models →