SE

SenseNova MARS 32B

SensenovaGGUF by mradermacherUpdated 2026-04-26
Download on Hugging Face
Parameters32.8B
Context256K
Layers64
Architectureqwen3vl
Smallest quant0.0 GB
Quick answer

SenseNova MARS 32B needs about 22.4 GB of memory at Q4_K_M (a 19.8 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 24 quants, from 0.0 GB to 26.9 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 8.6 GB.

SenseNova MARS 32B – 32.8B parameter vision‑language model

SenseNova MARS 32B is a 32.8B parameter vision‑language model built on the qwen3vl architecture. It supports interleaved visual reasoning and tool use, trained on a 262144 token context window. The model is released under the MIT license.

Good at

  • Tool‑enabled visual reasoning
  • Search‑driven image understanding
  • Fine‑grained visual tasks

Limits

  • No language list provided
  • Requires a vision backend
  • Large VRAM needed for full context

Best for

  • Visual search assistants
  • Fine‑grained image annotation
  • Knowledge‑intensive visual QA

Who it’s for. Developers building agentic vision‑language applications that need dynamic tool invocation and high‑resolution image reasoning. The model is suitable for research and production use where a large context window and tool support are required.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 2.7 GB
Fits Get file →
IQ1_S7.3 GB 10.0 GB
Too big Get file →
IQ1_M8.0 GB 10.6 GB
Too big Get file →
IQ2_XXS9.0 GB 11.7 GB
Too big Get file →
IQ2_XS10.0 GB 12.6 GB
Too big Get file →
IQ2_S10.5 GB 13.2 GB
Too big Get file →
IQ2_M11.4 GB 14.0 GB
Too big Get file →
Q2_K_S11.5 GB 14.1 GB
Too big Get file →
Q2_K12.3 GB 15.0 GB
Too big Get file →
IQ3_XXS12.8 GB 15.5 GB
Too big Get file →
IQ3_XS13.7 GB 16.4 GB
Too big Get file →
Q3_K_S14.4 GB 17.0 GB
Too big Get file →
IQ3_S14.4 GB 17.1 GB
Too big Get file →
IQ3_M14.9 GB 17.6 GB
Too big Get file →
Q3_K_M16.0 GB 18.6 GB
Too big Get file →
Q3_K_L17.3 GB 20.0 GB
Too big Get file →
IQ4_XS17.7 GB 20.3 GB
Too big Get file →
Q4_018.7 GB 21.4 GB
Too big Get file →
Q4_K_S18.8 GB 21.4 GB
Too big Get file →
Q4_K_M19.8 GB 22.4 GB
Too big Get file →
Q4_120.6 GB 23.3 GB
Too big Get file →
Q5_K_S22.6 GB 25.3 GB
Too big Get file →
Q5_K_M23.2 GB 25.9 GB
Too big Get file →
Q6_K26.9 GB 29.5 GB
Too big Get file →

For 8‑GB VRAM, the Q3_K_M quant (15.97 GB file, 18.6 GB memory at 8k context) is the smallest that still provides full functionality. If more VRAM is available, Q4_K_M (19.76 GB file, 22.4 GB memory) offers higher quality. For very limited memory, consider IQ1_S (7.32 GB file, 10 GB memory).

What GPU do you need for SenseNova MARS 32B?

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 the main file (0.0 GB file, about 2.7 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
  • 8 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 2.7 GB in use).
  • 12 GB VRAM: the best quant that fits is IQ2_XXS (9.0 GB file, about 11.7 GB in use).
  • 16 GB VRAM: the best quant that fits is IQ3_XXS (12.8 GB file, about 15.5 GB in use).
  • 24 GB VRAM: the best quant that fits is Q4_1 (20.6 GB file, about 23.3 GB in use).
  • 32 GB VRAM: the best quant that fits is Q6_K (26.9 GB file, about 29.5 GB in use).

Which SenseNova MARS 32B quant should you choose?

Q4_K_M is the usual starting point for SenseNova MARS 32B: it balances file size and quality well. Here it is a 19.8 GB file.

If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen3vl
Parameters32.8B
Layers64
Trained context262,144 tokens
KV cache (fp16)256 KB per token

Source

Quantized bymradermacher
FamilySensenova
Licensemit
Downloads (HF)561
Last updated2026-04-26

Frequently asked questions about SenseNova MARS 32B

How much VRAM does SenseNova MARS 32B need?

About 22.4 GB at Q4_K_M with an 8K context window. That is the 19.8 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 SenseNova MARS 32B on an 8 GB GPU?

Yes, with the smaller files: the 0.0 GB file needs about 2.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of SenseNova MARS 32B GGUF?

It depends on the quant: file is 0.0 GB, IQ1_S is 7.3 GB, IQ1_M is 8.0 GB, IQ2_XXS is 9.0 GB, IQ2_XS is 10.0 GB, IQ2_S is 10.5 GB and 18 more.

What context length does SenseNova MARS 32B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 2.1 GB at 8K and 8.6 GB at 32K.

Where can I download SenseNova MARS 32B GGUF?

The files are on Hugging Face in the mradermacher/SenseNova-MARS-32B-i1-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 SenseNova MARS 32B use?

Hugging Face lists the license as mit. 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 SenseNova MARS 32B on 8 GB VRAM?

The smallest quant that fits 8 GB VRAM is Q3_K_M, which requires 18.6 GB of memory at 8k context. It will not run fully on 8 GB VRAM without off‑loading.

Does the model support multiple languages?

No language list is provided; the model is released under MIT and has no explicit language restrictions.

Is there a vision backend required?

Yes, the model is a vision‑language model and requires a vision backend to process images.

Other GGUFs of SenseNova-MARS-32B

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Sensenova models

All Sensenova →

Other medium models

All medium models →