qwen3loop 0.6b sft deep supervision v1 needs about 3.5 GB of memory at Q8_0 (a 1.1 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 4 quants, from 0.6 GB (Q8_0) to 2.1 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 7.5 GB.
Qwen3Loop‑0.6B‑SFT‑Deep‑Supervision: 0.6B parameters, 56 logical layers for deep reasoning
Qwen3Loop‑0.6B‑SFT‑Deep‑Supervision is a 0.6B parameter model with 56 logical layers achieved through a looped architecture. It supports reasoning and vision tasks and was fine‑tuned on 6,324 curated samples. The model is released under the Apache‑2.0 license.
Good at
- Deep recursive reasoning
- Vision support
- Adaptive halting with 91.5% accuracy
Limits
- No vision support on some engines
- Requires 8k context for memory estimates
- Large VRAM needed for high‑precision quant
Best for
- Chat applications needing deep reasoning
- Vision‑enabled inference
- Fine‑tuned task adaptation
Who it’s for. Developers building local LLM services that require deep reasoning and vision capabilities within a small parameter footprint. The model is suitable for users who can run it on GPUs with sufficient VRAM and who need a lightweight yet powerful reasoning engine.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q8_0 | 0.6 GB | 3.0 GB | Fits | Get file → | |
| Q8_0 | 1.1 GB | 3.5 GB | Fits | Get file → | |
| F16 | 1.2 GB | 3.6 GB | Fits | Get file → | |
| F16 | 2.1 GB | 4.5 GB | Fits | Get file → |
For 8 GB VRAM, use the Q8_0 quantized file of 0.64 GB, which estimates 3 GB memory at 8k context. If more VRAM is available, the 1.11 GB Q8_0 file uses about 3.5 GB. For full precision, the 2.08 GB F16 file requires roughly 4.5 GB at 8k context. Estimates are approximate.
What GPU do you need for qwen3loop 0.6b sft deep supervision 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 F16 (1.2 GB file, about 3.6 GB in use).
- 6 GB VRAM: the best quant that fits is F16 (2.1 GB file, about 4.5 GB in use).
Which qwen3loop 0.6b sft deep supervision v1 quant should you choose?
Q8_0 (1.1 GB) is the middle option in this repository. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
If memory is tight, the smallest file is Q8_0 at 0.6 GB, with the quality trade-offs that come with fewer bits. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Frequently asked questions about qwen3loop 0.6b sft deep supervision v1
How much VRAM does qwen3loop 0.6b sft deep supervision v1 need?
About 3.5 GB at Q8_0 with an 8K context window. That is the 1.1 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 qwen3loop 0.6b sft deep supervision v1 on an 8 GB GPU?
Yes, with F16: the 2.1 GB file needs about 4.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of qwen3loop 0.6b sft deep supervision v1 GGUF?
It depends on the quant: Q8_0 is 0.6 GB, Q8_0 is 1.1 GB, F16 is 1.2 GB, F16 is 2.1 GB.
What context length does qwen3loop 0.6b sft deep supervision v1 support?
The GGUF header lists a trained context length of 40,960 tokens (40K). The KV cache grows with the context you use: about 1.9 GB at 8K and 7.5 GB at 32K.
Where can I download qwen3loop 0.6b sft deep supervision v1 GGUF?
The files are on Hugging Face in the Lordnyx/qwen3loop-0.6b-sft-deep-supervision-v1 repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does qwen3loop 0.6b sft deep supervision 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 this model on 8 GB VRAM?
Yes, the 0.64 GB Q8_0 quantized file uses about 3 GB memory at 8k context, fitting within 8 GB VRAM.
Does this model support vision tasks?
Yes, the model has detected vision abilities and includes a latent halting probe for adaptive depth.
What license applies to this model?
The model is released under the Apache‑2.0 license.