OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview needs about 16.7 GB of memory at Q4_K_M (a 15.8 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 3 quants, from 12.1 GB (MXFP4) to 41.8 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.
InternVL3_5-GPT-OSS-20B-A4B-Preview: 20.9B multimodal model in GGUF format
InternVL3_5-GPT-OSS-20B-A4B-Preview is a 20.9B parameter multimodal model based on GPT-OSS. It supports vision tasks and is available in several GGUF quantizations. The model is released under the apache-2.0 license.
Good at
- Vision‑language reasoning
- Dynamic visual token resolution
- Decoupled vision‑language deployment
Limits
- No vision support in some quantizations
- Large VRAM requirement for full precision
- Requires 131072 token context
Best for
- Image captioning with high detail
- Multimodal question answering
- Visual reasoning benchmarks
Who it’s for. Developers building multimodal applications who need an open‑source GPT‑OSS based model. The model is suitable for research and production use where vision and language are combined.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| MXFP4 | 12.1 GB | 13.0 GB | Too big | Get file → | |
| Q4_K_M | 15.8 GB | 16.7 GB | Too big | Get file → | |
| BF16 | 41.8 GB (4 parts) | 42.8 GB | Too big | Get file → |
For 8 GB VRAM, use the MXFP4 quant (12.1 GB file, 13 GB memory at 8 k context). For 16 GB VRAM, Q4_K_M (15.8 GB file, 16.7 GB memory) is suitable. For larger budgets, BF16 (41.85 GB file, 42.8 GB memory) provides the highest precision.
What GPU do you need for OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview?
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: the best quant that fits is MXFP4 (12.1 GB file, about 13.0 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_K_M (15.8 GB file, about 16.7 GB in use).
- 32 GB VRAM: the best quant that fits is Q4_K_M (15.8 GB file, about 16.7 GB in use).
- 48 GB VRAM: the best quant that fits is BF16 (41.8 GB file, about 42.8 GB in use).
Which OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview quant should you choose?
Q4_K_M is the usual starting point for OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview: it balances file size and quality well. Here it is a 15.8 GB file.
If memory is tight, the smallest file is MXFP4 at 12.1 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
Frequently asked questions about OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview
How much VRAM does OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview need?
About 16.7 GB at Q4_K_M with an 8K context window. That is the 15.8 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 OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 13.0 GB, so you would need a larger GPU or CPU offloading.
What is the file size of OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview GGUF?
It depends on the quant: MXFP4 is 12.1 GB, Q4_K_M is 15.8 GB, BF16 is 41.8 GB.
What context length does OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.
Where can I download OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview GGUF?
The files are on Hugging Face in the aviallon/InternVL3_5-GPT-OSS-20B-A4B-Preview-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 OpenGVLab/InternVL3_5-GPT-OSS-20B-A4B-Preview 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 InternVL3_5-GPT-OSS-20B-A4B-Preview on 8 GB VRAM?
Yes, the MXFP4 quantization requires 13 GB memory at 8 k context, which fits within an 8 GB VRAM setup with some overhead.
Does the model support vision tasks?
Yes, the model has detected vision abilities and is designed for multimodal use.
What license applies to this model?
The model is released under the apache-2.0 license.