LI

Ling 3.0 flash ROCmFP4 STRIX MTP

LingGGUF by raulvidisUpdated 2026-08-07
Download on Hugging Face
Parameters127B
Context256K
Layers43
Architecturebailingmoe3
Smallest quant69.7 GB
Quick answer

Ling 3.0 flash ROCmFP4 STRIX MTP needs about 94.7 GB of memory at Q4_0 (a 69.7 GB file) with an 8K context window, which is more than a 48 GB GPU holds. Longer context needs more: at 32K tokens the KV cache alone adds about 98.1 GB.

Ling 3.0 flash ROCmFP4 STRIX MTP – 127B hybrid MoE model for AMD Strix Halo

Ling 3.0 flash is a 127B parameter hybrid linear model with 5.1B active parameters. It is quantized to Q4_0 for AMD ROCmFP4 and supports MTP speculative decoding and SwiGLU clamps. The model is trained on 262144 context tokens and released under MIT.

Good at

  • Long‑context reasoning
  • Agentic workflows
  • Efficient AMD inference

Limits

  • Requires ROCmFPX llama.cpp fork
  • No vision support
  • Large VRAM needed

Best for

  • Production agent pipelines
  • Long‑horizon coding tasks
  • Complex research workflows

Who it’s for. Developers building production‑grade agents on AMD GPUs. The model is suited for teams needing efficient long‑context inference and advanced speculative decoding on Strix Halo hardware.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_069.7 GB (2 parts) 94.7 GB
Too big Get file →

For 8 GB VRAM, the Q4_0 quant file is 69.7 GB and requires ~94.7 GB of RAM at 8 k context – not feasible. A smaller quant such as Q4_0 is the only available option; memory estimates are approximate. Split the model into its two shards for loading.

What GPU do you need for Ling 3.0 flash ROCmFP4 STRIX MTP?

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: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: no quant fits at an 8K context.
  • 32 GB VRAM: no quant fits at an 8K context.
  • 48 GB VRAM: no quant fits at an 8K context.

Which Ling 3.0 flash ROCmFP4 STRIX MTP quant should you choose?

Q4_0 (69.7 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturebailingmoe3
Parameters127B (approximate, mixture of experts)
Layers43
Trained context262,144 tokens
KV cache (fp16)2924 KB per token

Source

Quantized byraulvidis
FamilyLing
Licensemit
Downloads (HF)92
Last updated2026-08-07

Frequently asked questions about Ling 3.0 flash ROCmFP4 STRIX MTP

How much VRAM does Ling 3.0 flash ROCmFP4 STRIX MTP need?

About 94.7 GB at Q4_0 with an 8K context window. That is the 69.7 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 Ling 3.0 flash ROCmFP4 STRIX MTP on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 94.7 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Ling 3.0 flash ROCmFP4 STRIX MTP GGUF?

It depends on the quant: Q4_0 is 69.7 GB.

What context length does Ling 3.0 flash ROCmFP4 STRIX MTP support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 24.5 GB at 8K and 98.1 GB at 32K.

Where can I download Ling 3.0 flash ROCmFP4 STRIX MTP GGUF?

The files are on Hugging Face in the raulvidis/Ling-3.0-flash-ROCmFP4-STRIX-MTP-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 Ling 3.0 flash ROCmFP4 STRIX MTP 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 this model on 8 GB VRAM?

No. The Q4_0 quant requires ~94.7 GB of RAM at 8 k context, far beyond 8 GB VRAM.

Does this model support vision?

No, the model has no vision capabilities.

What hardware is needed?

It requires an AMD Strix Halo GPU with ROCmFP4 kernels and the ROCmFPX llama.cpp fork.

Other GGUFs of Ling-3.0-flash

Similar models

Browse all →

More from raulvidis

All by raulvidis →

More Ling models

All Ling →

Other large models

All large models →