NA

Nanbeige 16B Chat 32K

NanbeigeGGUF by TheBlokeUpdated 2023-11-18
Download on Hugging Face
Parameters15.8B
Context4K
Layers48
Architecturellama
Smallest quant6.6 GB
Quick answer

Nanbeige 16B Chat 32K needs about 18.1 GB of memory at Q4_K_M (a 9.6 GB file) with an 8K context window, so it fits on a 24 GB GPU. This repository has 13 quants, from 6.6 GB (Q2_K) to 31.7 GB.

Nanbeige 16B Chat 32K – 15.8B parameter llama chat model

Nanbeige 16B Chat 32K is a 15.8B parameter llama architecture model trained on 4096 token context. It is available in multiple GGUF quantisations for CPU and GPU inference. The model is licensed under apache-2.0.

Good at

  • Coding assistance
  • Chat support
  • Contextual reasoning

Limits

  • No vision support
  • Context limited to 4096 tokens
  • Requires GPU for best performance

Best for

  • Local chatbots
  • Code generation tools
  • Educational tutoring

Who it’s for. Developers and researchers who need a large llama model for local inference. The model is suitable for those who can run GGUF files on llama.cpp or compatible libraries.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K6.6 GB 11.2 GB
Too big Get file →
Q3_K_S6.9 GB 11.5 GB
Too big Get file →
Q3_K_M7.7 GB 12.3 GB
Too big Get file →
Q3_K_L8.5 GB 13.0 GB
Too big Get file →
Q4_09.0 GB 13.5 GB
Too big Get file →
Q4_K_S9.0 GB 13.6 GB
Too big Get file →
Q4_K_M9.6 GB 14.1 GB
Too big Get file →
Q5_010.9 GB 15.5 GB
Too big Get file →
Q5_K_S10.9 GB 15.5 GB
Too big Get file →
Q5_K_M11.2 GB 15.8 GB
Too big Get file →
Q6_K13.0 GB 17.5 GB
Too big Get file →
Q8_017.1 GB 21.6 GB
Too big Get file →
—31.7 GB 36.2 GB
Too big Get file →

For 8 GB VRAM, use Q4_0 (file size 8.99 GB, estimated 17.5 GB memory at 8 k context). For 12 GB VRAM, Q4_K_M (9.59 GB, 18.1 GB memory) is suitable. For 16 GB VRAM, Q5_K_M (11.24 GB, 19.8 GB memory) offers higher quality.

What GPU do you need for Nanbeige 16B Chat 32K?

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: the best quant that fits is Q3_K_S (6.9 GB file, about 15.5 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (13.0 GB file, about 21.5 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (17.1 GB file, about 25.7 GB in use).
  • 48 GB VRAM: the best quant that fits is the main file (31.7 GB file, about 40.2 GB in use).

Which Nanbeige 16B Chat 32K quant should you choose?

Q4_K_M is the usual starting point for Nanbeige 16B Chat 32K: it balances file size and quality well. Here it is a 9.6 GB file.

The Q8_0 file is 17.1 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 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

Architecturellama
Parameters15.8B
Layers48
Trained context4,096 tokens
KV cache (fp16)960 KB per token

Source

Quantized byTheBloke
FamilyNanbeige
Licenseapache-2.0
Downloads (HF)968
Last updated2023-11-18

Frequently asked questions about Nanbeige 16B Chat 32K

How much VRAM does Nanbeige 16B Chat 32K need?

About 18.1 GB at Q4_K_M with an 8K context window. That is the 9.6 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 Nanbeige 16B Chat 32K on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 15.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Nanbeige 16B Chat 32K GGUF?

It depends on the quant: Q2_K is 6.6 GB, Q3_K_S is 6.9 GB, Q3_K_M is 7.7 GB, Q3_K_L is 8.5 GB, Q4_0 is 9.0 GB, Q4_K_S is 9.0 GB and 7 more.

What context length does Nanbeige 16B Chat 32K support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 8.1 GB at 8K and 4.0 GB at 4K.

Where can I download Nanbeige 16B Chat 32K GGUF?

The files are on Hugging Face in the TheBloke/Nanbeige-16B-Chat-32K-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 Nanbeige 16B Chat 32K 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 Nanbeige 16B Chat 32K on 8 GB VRAM?

Yes, use the Q4_0 quantised model which requires about 17.5 GB memory at 8 k context, which fits within 8 GB VRAM when using GPU acceleration.

What license does the model use?

The model is licensed under apache-2.0, allowing free use and modification.

Does the model support vision tasks?

No, the model has no vision capabilities and is designed for text-only tasks.

Similar models

Browse all →

More from TheBloke

All by TheBloke →

More Nanbeige models

All Nanbeige →

Other medium models

All medium models →