NA

Narwhal 7b

NarwhalGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters6.74B
Context2K
Layers32
Architecturellama
Smallest quant2.5 GB
Quick answer

Narwhal 7b needs about 8.9 GB of memory at Q4_K_M (a 4.1 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 14 quants, from 2.5 GB (Q2_K) to 7.2 GB (Q8_0).

Narwhal 7b – 6.74B parameter Llama model

Narwhal 7b is a 6.74B parameter Llama model built from a blend of Stable Beluga, MegaCoder, Wizard‑Math, and Llama Chat7b. It excels at mathematical tasks and general query response. The model is not intended for commercial use due to underlying licenses.

Good at

  • Mathematical problem solving
  • General question answering
  • Prompted Llama v2 style responses

Limits

  • No vision support
  • Not for commercial use
  • Context limited to 2048 tokens

Best for

  • Educational math tutoring
  • Research question answering
  • Non‑commercial prototyping

Who it’s for. Developers and researchers who need a capable, non‑commercial Llama model for math and general text tasks. The model is suitable for local inference with GGUF and llama.cpp.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K2.5 GB 4.1 GB
Fits Get file →
IQ3_XS2.8 GB 4.4 GB
Fits Get file →
IQ3_S2.9 GB 4.5 GB
Fits Get file →
Q3_K_S2.9 GB 4.5 GB
Fits Get file →
IQ3_M3.1 GB 4.7 GB
Fits Get file →
Q3_K_M3.3 GB 4.9 GB
Fits Get file →
Q3_K_L3.6 GB 5.2 GB
Fits Get file →
IQ4_XS3.6 GB 5.2 GB
Fits Get file →
Q4_K_S3.9 GB 5.4 GB
Fits Get file →
Q4_K_M4.1 GB 5.7 GB
Fits Get file →
Q5_K_S4.7 GB 6.2 GB
Fits Get file →
Q5_K_M4.8 GB 6.4 GB
Fits Get file →
Q6_K5.5 GB 7.1 GB
Fits Get file →
Q8_07.2 GB 8.7 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (est. 8.9 GB at 8 k context) or Q5_K_S (est. 9.4 GB). For 12 GB VRAM, Q8_0 (est. 12 GB) offers the highest quality. Smaller budgets can use Q3_K_S or IQ3_S, which need about 7.7 GB.

What GPU do you need for Narwhal 7b?

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 IQ3_M (3.1 GB file, about 7.9 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (7.2 GB file, about 12.0 GB in use).

Which Narwhal 7b quant should you choose?

Q4_K_M is the usual starting point for Narwhal 7b: it balances file size and quality well. Here it is a 4.1 GB file.

The Q8_0 file is 7.2 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 2.5 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
Parameters6.74B
Layers32
Trained context2,048 tokens
KV cache (fp16)512 KB per token

Source

Quantized bymradermacher
FamilyNarwhal
Downloads (HF)1,823
Last updated2024-05-06

Frequently asked questions about Narwhal 7b

How much VRAM does Narwhal 7b need?

About 8.9 GB at Q4_K_M with an 8K context window. That is the 4.1 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 Narwhal 7b on an 8 GB GPU?

Yes, with IQ3_M: the 3.1 GB file needs about 7.9 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Narwhal 7b GGUF?

It depends on the quant: Q2_K is 2.5 GB, IQ3_XS is 2.8 GB, IQ3_S is 2.9 GB, Q3_K_S is 2.9 GB, IQ3_M is 3.1 GB, Q3_K_M is 3.3 GB and 8 more.

What context length does Narwhal 7b support?

The GGUF header lists a trained context length of 2,048 tokens (2K). The KV cache grows with the context you use: about 4.3 GB at 8K and 1.1 GB at 2K.

Where can I download Narwhal 7b GGUF?

The files are on Hugging Face in the mradermacher/Narwhal-7b-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 Narwhal 7b on 8 GB VRAM?

Yes, with Q4_K_M or Q5_K_S quants, which estimate 8.9 GB or 9.4 GB at 8 k context. Use a lower context length to reduce memory.

Is Narwhal 7b suitable for commercial projects?

No, the model is not intended for commercial use due to licenses of underlying datasets.

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Narwhal models

All Narwhal →

Other small models

All small models →