SA

SambaLingo Arabic Chat 70B

SambalingoGGUF by mradermacherUpdated 2024-12-14
Download on Hugging Face
Parameters69.4B
Context4K
Layers80
Architecturellama
Smallest quant14.7 GB
Quick answer

SambaLingo Arabic Chat 70B needs about 44.9 GB of memory at Q4_K_M (a 41.7 GB file) with an 8K context window, so it fits on a 48 GB GPU. This repository has 21 quants, from 14.7 GB (IQ1_S) to 49.1 GB (Q5_K_M).

SambaLingo Arabic Chat 70B – 69.4B parameter language model

SambaLingo Arabic Chat 70B is a 69.4B parameter chat model trained on Arabic and English. It is fine‑tuned from Llama‑2‑70b using direct preference optimization. The model supports a 4096 token context window.

Good at

  • Arabic and English chat
  • Human‑aligned responses
  • Large context handling

Limits

  • Requires large VRAM – smallest quant needs ~17.9GB at 8k context
  • No vision support
  • License: llama2

Best for

  • Chatbots in Arabic
  • Multilingual customer support
  • Content generation in Arabic and English

Who it’s for. Developers building Arabic‑English conversational agents who need a large, fine‑tuned language model. The model is suitable for research and production use where a 70B parameter model is required.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S14.7 GB 16.6 GB
Too big Get file →
IQ1_M16.2 GB 18.0 GB
Too big Get file →
IQ2_XXS18.5 GB 20.3 GB
Too big Get file →
IQ2_XS20.5 GB 22.4 GB
Too big Get file →
IQ2_S21.6 GB 23.4 GB
Too big Get file →
IQ2_M23.5 GB 25.3 GB
Too big Get file →
Q2_K_S23.8 GB 25.6 GB
Too big Get file →
Q2_K25.7 GB 27.5 GB
Too big Get file →
IQ3_XXS26.8 GB 28.7 GB
Too big Get file →
IQ3_XS28.6 GB 30.4 GB
Too big Get file →
IQ3_S30.2 GB 32.0 GB
Too big Get file →
Q3_K_S30.2 GB 32.0 GB
Too big Get file →
IQ3_M31.2 GB 33.0 GB
Too big Get file →
Q3_K_M33.5 GB 35.4 GB
Too big Get file →
Q3_K_L36.4 GB 38.3 GB
Too big Get file →
IQ4_XS37.1 GB 39.0 GB
Too big Get file →
Q4_039.3 GB 41.1 GB
Too big Get file →
Q4_K_S39.5 GB 41.4 GB
Too big Get file →
Q4_K_M41.7 GB 43.6 GB
Too big Get file →
Q5_K_S47.8 GB 49.6 GB
Too big Get file →
Q5_K_M49.1 GB 50.9 GB
Too big Get file →

The smallest available quant is IQ1_S (14.75 GB file) and needs about 17.9 GB VRAM at 8k context. For 16 GB VRAM use IQ1_M (16.15 GB file, ~19.3 GB VRAM). For 32 GB VRAM or higher, Q4_K_M (41.71 GB file, ~44.9 GB VRAM) is a good balance between size and memory usage.

What GPU do you need for SambaLingo Arabic Chat 70B?

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: the best quant that fits is IQ2_XS (20.5 GB file, about 23.7 GB in use).
  • 32 GB VRAM: the best quant that fits is IQ3_XS (28.6 GB file, about 31.8 GB in use).
  • 48 GB VRAM: the best quant that fits is Q4_K_M (41.7 GB file, about 44.9 GB in use).

Which SambaLingo Arabic Chat 70B quant should you choose?

Q4_K_M is the usual starting point for SambaLingo Arabic Chat 70B: it balances file size and quality well. Here it is a 41.7 GB file.

If memory is tight, the smallest file is IQ1_S at 14.7 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
Parameters69.4B
Layers80
Trained context4,096 tokens
KV cache (fp16)320 KB per token

Source

Quantized bymradermacher
Licensellama2
Downloads (HF)678
Last updated2024-12-14

Frequently asked questions about SambaLingo Arabic Chat 70B

How much VRAM does SambaLingo Arabic Chat 70B need?

About 44.9 GB at Q4_K_M with an 8K context window. That is the 41.7 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 SambaLingo Arabic Chat 70B on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 17.9 GB, so you would need a larger GPU or CPU offloading.

What is the file size of SambaLingo Arabic Chat 70B GGUF?

It depends on the quant: IQ1_S is 14.7 GB, IQ1_M is 16.2 GB, IQ2_XXS is 18.5 GB, IQ2_XS is 20.5 GB, IQ2_S is 21.6 GB, IQ2_M is 23.5 GB and 15 more.

What context length does SambaLingo Arabic Chat 70B support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 2.7 GB at 8K and 1.3 GB at 4K.

Where can I download SambaLingo Arabic Chat 70B GGUF?

The files are on Hugging Face in the mradermacher/SambaLingo-Arabic-Chat-70B-i1-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 SambaLingo Arabic Chat 70B use?

Hugging Face lists the license as llama2. 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 smallest quant requires about 17.9 GB VRAM at 8k context.

Does this model support vision?

No, it is a language model only.

What license does it use?

The model is released under the llama2 license.

Other GGUFs of SambaLingo-Arabic-Chat-70B

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Sambalingo models

All Sambalingo →

Other large models

All large models →