KI

Kimi VL A3B Thinking 2506

KimiGGUF by mradermacherUpdated 2025-09-17
Download on Hugging Face
Parameters16B
Context128K
Layers27
Architecturedeepseek2
Smallest quant6.6 GB
Quick answer

Kimi VL A3B Thinking 2506 needs about 11.5 GB of memory at Q4_K_M (a 10.5 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 11 quants, from 6.6 GB (Q2_K) to 17.0 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

Kimi VL A3B Thinking 2506 – 16B multimodal LLM with reasoning and vision

Kimi VL A3B Thinking 2506 is a 16B deepseek2 model with 3B active parameters. It supports 131072 context tokens and excels in multimodal reasoning, vision, and video understanding. The model is released under the MIT license.

Good at

  • Multimodal reasoning
  • Vision perception
  • Video understanding

Limits

  • Large VRAM requirement for full context
  • No explicit language support list
  • Requires flash‑attn for efficient inference

Best for

  • High‑resolution image analysis
  • Video scenario reasoning
  • Multimodal question answering

Who it’s for. Developers building multimodal applications that need reasoning and vision capabilities. Researchers interested in open‑source large models with a 16B parameter base.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K6.6 GB 7.6 GB
Tight Get file →
Q3_K_S7.6 GB 8.6 GB
Too big Get file →
Q3_K_M8.3 GB 9.3 GB
Too big Get file →
Q3_K_L8.6 GB 9.6 GB
Too big Get file →
IQ4_XS8.8 GB 9.8 GB
Too big Get file →
Q4_K_S9.7 GB 10.7 GB
Too big Get file →
Q4_K_M10.5 GB 11.5 GB
Too big Get file →
Q5_K_S11.3 GB 12.3 GB
Too big Get file →
Q5_K_M12.0 GB 13.0 GB
Too big Get file →
Q6_K14.3 GB 15.3 GB
Too big Get file →
Q8_017.0 GB 18.0 GB
Too big Get file →

For 8GB VRAM, use Q4_K_S (file size 9.71 GB, memory 10.7 GB at 8K context). For 12GB VRAM, Q5_K_S (11.34 GB, 12.3 GB). For 16GB VRAM, Q6_K (14.28 GB, 15.3 GB). Estimates are approximate.

What GPU do you need for Kimi VL A3B Thinking 2506?

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: the best quant that fits is Q2_K (6.6 GB file, about 7.6 GB in use).
  • 12 GB VRAM: the best quant that fits is Q4_K_M (10.5 GB file, about 11.5 GB in use).
  • 16 GB VRAM: the best quant that fits is Q6_K (14.3 GB file, about 15.3 GB in use).
  • 24 GB VRAM: the best quant that fits is Q8_0 (17.0 GB file, about 18.0 GB in use).

Which Kimi VL A3B Thinking 2506 quant should you choose?

Q4_K_M is the usual starting point for Kimi VL A3B Thinking 2506: it balances file size and quality well. Here it is a 10.5 GB file.

The Q8_0 file is 17.0 GB, about 1.6 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 6.6 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

Architecturedeepseek2
Parameters16B (approximate, mixture of experts)
Active parameters3B
Layers27
Trained context131,072 tokens
KV cache (fp16)57 KB per token

Source

Quantized bymradermacher
FamilyKimi
Licensemit
Downloads (HF)835
Last updated2025-09-17

Frequently asked questions about Kimi VL A3B Thinking 2506

How much VRAM does Kimi VL A3B Thinking 2506 need?

About 11.5 GB at Q4_K_M with an 8K context window. That is the 10.5 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 Kimi VL A3B Thinking 2506 on an 8 GB GPU?

Yes, with Q2_K: the 6.6 GB file needs about 7.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Kimi VL A3B Thinking 2506 GGUF?

It depends on the quant: Q2_K is 6.6 GB, Q3_K_S is 7.6 GB, Q3_K_M is 8.3 GB, Q3_K_L is 8.6 GB, IQ4_XS is 8.8 GB, Q4_K_S is 9.7 GB and 5 more.

What context length does Kimi VL A3B Thinking 2506 support?

The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.

Where can I download Kimi VL A3B Thinking 2506 GGUF?

The files are on Hugging Face in the mradermacher/Kimi-VL-A3B-Thinking-2506-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 Kimi VL A3B Thinking 2506 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 Kimi VL A3B Thinking 2506 on 8GB VRAM?

Use the Q4_K_S quant, which requires about 10.7 GB of VRAM at 8K context. It may still fit with optimizations.

Does the model support video input?

Yes, it extends to video scenarios and performs well on VideoMMMU and Video-MME benchmarks.

What license applies to this model?

The model is released under the MIT license.

Other GGUFs of Kimi-VL-A3B-Thinking-2506

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Kimi models

All Kimi →

Other medium models

All medium models →