LL

Llama 30b

LlamaGGUF by mradermacherUpdated 2024-09-11
Download on Hugging Face
Parameters32.5B
Context2K
Layers60
Architecturellama
Smallest quant12.0 GB
Quick answer

Llama 30b needs about 33.2 GB of memory at Q4_K_M (a 19.6 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 14 quants, from 12.0 GB (Q2_K) to 34.6 GB (Q8_0).

Llama 30b SuperCOT quantized model

Llama 30b SuperCOT is a 32.5 billion‑parameter Llama model fine‑tuned with SuperCOT LoRA. It is available in multiple GGUF quantizations ranging from Q2_K to Q8_0. The model supports 2048‑token context length.

Good at

  • Fine‑tuned SuperCOT prompts
  • Large‑parameter reasoning
  • LoRA compatibility

Limits

  • No vision support
  • License not specified
  • Requires 8 k context for memory estimates

Best for

  • Local LLM notebooks
  • LangChain integration
  • Prompt engineering experiments

Who it’s for. Developers and researchers who need a large Llama model with SuperCOT fine‑tuning for local deployment. The model is suitable for those using llama.cpp or Oobabooga to run GGUF files.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K12.0 GB 15.8 GB
Too big Get file →
IQ3_XS13.3 GB 17.1 GB
Too big Get file →
IQ3_S14.1 GB 17.8 GB
Too big Get file →
Q3_K_S14.1 GB 17.8 GB
Too big Get file →
IQ3_M14.9 GB 18.7 GB
Too big Get file →
Q3_K_M15.8 GB 19.5 GB
Too big Get file →
Q3_K_L17.3 GB 21.1 GB
Too big Get file →
IQ4_XS17.5 GB 21.2 GB
Too big Get file →
Q4_K_S18.5 GB 22.3 GB
Too big Get file →
Q4_K_M19.6 GB 23.4 GB
Too big Get file →
Q5_K_S22.4 GB 26.2 GB
Too big Get file →
Q5_K_M23.0 GB 26.8 GB
Too big Get file →
Q6_K26.7 GB 30.5 GB
Too big Get file →
Q8_034.6 GB 38.3 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_S (file size 18.48 GB, estimated 32.1 GB memory at 8 k context). For 16 GB VRAM, Q5_K_M (23.05 GB, 36.6 GB memory). For 32 GB VRAM, Q6_K (26.69 GB, 40.3 GB memory). Estimates are approximate.

What GPU do you need for Llama 30b?

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: the best quant that fits is IQ4_XS (17.5 GB file, about 31.1 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (26.7 GB file, about 40.3 GB in use).

Which Llama 30b quant should you choose?

Q4_K_M is the usual starting point for Llama 30b: it balances file size and quality well. Here it is a 19.6 GB file.

The Q8_0 file is 34.6 GB, about 1.8 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 12.0 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

Architecturellama
Parameters32.5B
Layers60
Trained context2,048 tokens
KV cache (fp16)1560 KB per token

Source

Quantized bymradermacher
FamilyLlama
Downloads (HF)1,198
Last updated2024-09-11

Frequently asked questions about Llama 30b

How much VRAM does Llama 30b need?

About 33.2 GB at Q4_K_M with an 8K context window. That is the 19.6 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 Llama 30b on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 25.6 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Llama 30b GGUF?

It depends on the quant: Q2_K is 12.0 GB, IQ3_XS is 13.3 GB, IQ3_S is 14.1 GB, Q3_K_S is 14.1 GB, IQ3_M is 14.9 GB, Q3_K_M is 15.8 GB and 8 more.

What context length does Llama 30b support?

The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 13.1 GB at 8K and 3.3 GB at 2K.

Where can I download Llama 30b GGUF?

The files are on Hugging Face in the mradermacher/llama-30b-supercot-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 this model on 8 GB VRAM?

Use the Q4_K_S quantization; it requires about 32.1 GB memory at 8 k context, which is beyond 8 GB VRAM, so it is not feasible.

Does this model support vision tasks?

No, the model has no vision capabilities.

What is the context length of this model?

The model supports 2048 tokens of context.

Other GGUFs of llama-30b-supercot

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Llama models

All Llama →

Other medium models

All medium models →