SA

SAVER 4B

SaverGGUF by mradermacherUpdated 2026-10-08
Download on Hugging Face
Parameters4.84B
Context256K
Layers32
Architectureqwen35
Smallest quant2.1 GB
Quick answer

SAVER 4B needs about 4.6 GB of memory at Q4_K_M (a 3.1 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 12 quants, from 2.1 GB (Q2_K) to 9.7 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

SAVER 4B – sparse video reasoning model

SAVER 4B is a reinforcement‑learning post‑trained model based on Qwen3.5‑4B. It focuses on sparse video question answering and temporal grounding. The model is released in GGUF format with multiple quantizations.

Good at

  • Sparse video reasoning
  • Temporal grounding
  • Low‑frame inference

Limits

  • No vision support in GGUF
  • License not specified
  • No large‑context support beyond 262144 tokens

Best for

  • Video QA with limited frames
  • Temporal grounding tasks
  • Efficient inference on low‑VRAM GPUs

Who it’s for. Developers building video‑language applications that need efficient inference. Researchers interested in sparse video reasoning and reinforcement post‑training.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K2.1 GB 3.7 GB
Fits Get file →
Q3_K_S2.3 GB 3.9 GB
Fits Get file →
Q3_K_M2.5 GB 4.1 GB
Fits Get file →
Q3_K_L2.7 GB 4.3 GB
Fits Get file →
IQ4_XS2.9 GB 4.4 GB
Fits Get file →
Q4_K_S2.9 GB 4.5 GB
Fits Get file →
Q4_K_M3.1 GB 4.6 GB
Fits Get file →
Q5_K_S3.4 GB 5.0 GB
Fits Get file →
Q5_K_M3.5 GB 5.1 GB
Fits Get file →
Q6_K4.0 GB 5.6 GB
Fits Get file →
Q8_05.2 GB 6.7 GB
Fits Get file →
F169.7 GB 11.3 GB
Too big Get file →

For 8 GB VRAM use Q4_K_S (file size 2.92 GB, memory 4.5 GB). For 12 GB VRAM use Q5_K_S (file size 3.43 GB, memory 5 GB). For 16 GB VRAM use Q6_K (file size 3.99 GB, memory 5.6 GB). Estimates are approximate.

What GPU do you need for SAVER 4B?

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 Q3_K_S (2.3 GB file, about 3.9 GB in use).
  • 6 GB VRAM: the best quant that fits is Q6_K (4.0 GB file, about 5.6 GB in use).
  • 8 GB VRAM: the best quant that fits is Q8_0 (5.2 GB file, about 6.7 GB in use).
  • 12 GB VRAM: the best quant that fits is F16 (9.7 GB file, about 11.3 GB in use).

Which SAVER 4B quant should you choose?

Q4_K_M is the usual starting point for SAVER 4B: it balances file size and quality well. Here it is a 3.1 GB file.

The Q8_0 file is 5.2 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 2.1 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

Architectureqwen35
Parameters4.84B
Layers32
Trained context262,144 tokens
KV cache (fp16)128 KB per token

Source

Quantized bymradermacher
FamilySaver
Base modellmsdss/SAVER-4B
Downloads (HF)737
Last updated2026-10-08

Frequently asked questions about SAVER 4B

How much VRAM does SAVER 4B need?

About 4.6 GB at Q4_K_M with an 8K context window. That is the 3.1 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 SAVER 4B on an 8 GB GPU?

Yes, with Q8_0: the 5.2 GB file needs about 6.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of SAVER 4B GGUF?

It depends on the quant: Q2_K is 2.1 GB, Q3_K_S is 2.3 GB, Q3_K_M is 2.5 GB, Q3_K_L is 2.7 GB, IQ4_XS is 2.9 GB, Q4_K_S is 2.9 GB and 6 more.

What context length does SAVER 4B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download SAVER 4B GGUF?

The files are on Hugging Face in the mradermacher/SAVER-4B-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

Can I run SAVER 4B on 8 GB VRAM?

Yes, use the Q4_K_S quantization which requires about 4.5 GB of VRAM at 8 k context.

Does the model support vision tasks?

The GGUF release does not include vision support; it is designed for sparse video reasoning.

What license applies to this model?

The base model is under Apache 2.0; the license for the GGUF quantizations is not specified.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Saver models

All Saver →

Other small models

All small models →