QW

MiniMax-H3-Prompt-Rewriter-8B

QwenGGUF by indhic-aiUpdated 2026-08-19
Download on Hugging Face
Parameters8.19B
Context256K
Layers36
Architectureqwen3vl
Smallest quant8.7 GB
Quick answer

MiniMax-H3-Prompt-Rewriter-8B needs about 10.4 GB of memory at Q8_0 (a 8.7 GB file) with an 8K context window, so it fits on a 12 GB GPU. Longer context needs more: at 32K tokens the KV cache alone adds about 4.8 GB.

MiniMax-H3 Prompt Rewriter 8B – LoRA‑merged GGUF for vision‑language tasks

MiniMax-H3 Prompt Rewriter 8B is a Qwen3‑VL‑8B‑Instruct base model merged with a LoRA adapter. It rewrites short user requests into production‑ready prompts for audio‑video generation tasks. The model supports vision and tool usage within llama.cpp.

Good at

  • Vision‑language prompt rewriting
  • Audio‑video generation prep
  • Tool invocation

Limits

  • No vision support without mmproj file
  • Requires 8.71 GB GGUF file
  • Memory estimate 10.4 GB at 8 k context

Best for

  • Generating prompts for MiniMax‑H3 audio‑video models
  • Preparing T2VA, I2VA, L2VA, FL2VA tasks
  • Automating prompt creation in local LLM workflows

Who it’s for. Developers and researchers who need a lightweight prompt rewriter for MiniMax‑H3 audio‑video generation. The model is suitable for local deployment with llama.cpp and requires a GPU with at least 8 GB VRAM plus system RAM for context handling.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q8_08.7 GB 10.4 GB
Too big Get file →

The only available quant is Q8_0, 8.71 GB in size. It needs about 10.4 GB of VRAM at 8 k context. For lower VRAM budgets, no other quant is available, so the model cannot be run with smaller memory footprints.

What GPU do you need for MiniMax-H3-Prompt-Rewriter-8B?

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: the best quant that fits is Q8_0 (8.7 GB file, about 10.4 GB in use).

Which MiniMax-H3-Prompt-Rewriter-8B quant should you choose?

Q8_0 (8.7 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.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen3vl
Parameters8.19B
Layers36
Trained context262,144 tokens
KV cache (fp16)144 KB per token

Source

Quantized byindhic-ai
FamilyQwen
Licenseapache-2.0
Downloads (HF)616
Last updated2026-08-19

Frequently asked questions about MiniMax-H3-Prompt-Rewriter-8B

How much VRAM does MiniMax-H3-Prompt-Rewriter-8B need?

About 10.4 GB at Q8_0 with an 8K context window. That is the 8.7 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 MiniMax-H3-Prompt-Rewriter-8B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.4 GB, so you would need a larger GPU or CPU offloading.

What is the file size of MiniMax-H3-Prompt-Rewriter-8B GGUF?

It depends on the quant: Q8_0 is 8.7 GB.

What context length does MiniMax-H3-Prompt-Rewriter-8B support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.2 GB at 8K and 4.8 GB at 32K.

Where can I download MiniMax-H3-Prompt-Rewriter-8B GGUF?

The files are on Hugging Face in the indhic-ai/MiniMax_H3-Prompt_Rewriter-8B-LORA-Merged-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 MiniMax-H3-Prompt-Rewriter-8B 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 MiniMax-H3 Prompt Rewriter on 8 GB VRAM?

The Q8_0 quant requires 10.4 GB VRAM at 8 k context, so 8 GB is insufficient.

Does this model support vision tasks?

Yes, but only when the mmproj file is present for image‑conditioned tasks.

What license applies to this model?

The model is released under the apache-2.0 license.

Other GGUFs of Qwen3-VL-8B-Instruct

Similar models

Browse all →

More Qwen models

All Qwen →

Other small models

All small models →