SA

Sailor 4B

SailorGGUF by RichardErkhovUpdated 2024-08-19
Download on Hugging Face
Parameters3.95B
Context32K
Layers40
Architectureqwen2
Smallest quant1.6 GB
Quick answer

Sailor 4B needs about 6.3 GB of memory at Q4_K_M (a 2.5 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 22 quants, from 1.6 GB (Q2_K) to 4.2 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 13.4 GB.

Sailor‑4B: 3.95B‑parameter multilingual Qwen2 model

Sailor‑4B is a 3.95 billion‑parameter Qwen2 model trained on multilingual datasets. It supports a 32768‑token context window and is released under the apache‑2.0 license.

Good at

  • Multilingual text generation
  • Large context handling
  • Low‑memory inference

Limits

  • No vision support
  • Requires at least 5.5 GB VRAM for Q2_K
  • No mixed‑precision training

Best for

  • Chatbots needing 32k context
  • Fine‑tuning on small datasets
  • Embedded inference on edge devices

Who it’s for. Developers building multilingual LLM applications who need a compact model with a large context window. The model is suitable for local deployment and research projects that require an open‑source license.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K1.6 GB 5.5 GB
Fits Get file →
IQ3_XS1.8 GB 5.6 GB
Fits Get file →
IQ3_S1.9 GB 5.7 GB
Fits Get file →
Q3_K_S1.9 GB 5.7 GB
Fits Get file →
IQ3_M1.9 GB 5.8 GB
Fits Get file →
Q3_K2.0 GB 5.9 GB
Fits Get file →
Q3_K_M2.0 GB 5.9 GB
Fits Get file →
Q3_K_L2.2 GB 6.0 GB
Fits Get file →
IQ4_XS2.2 GB 6.1 GB
Fits Get file →
Q4_02.3 GB 6.2 GB
Fits Get file →
IQ4_NL2.3 GB 6.2 GB
Fits Get file →
Q4_K_S2.3 GB 6.2 GB
Fits Get file →
Q4_K2.5 GB 6.3 GB
Fits Get file →
Q4_K_M2.5 GB 6.3 GB
Fits Get file →
Q4_12.6 GB 6.4 GB
Fits Get file →
Q5_02.8 GB 6.6 GB
Fits Get file →
Q5_K_S2.8 GB 6.6 GB
Fits Get file →
Q5_K2.8 GB 6.7 GB
Fits Get file →
Q5_K_M2.8 GB 6.7 GB
Fits Get file →
Q5_13.0 GB 6.9 GB
Fits Get file →
Q6_K3.2 GB 7.1 GB
Fits Get file →
Q8_04.2 GB 8.1 GB
Too big Get file →

For 8 GB VRAM, use Q4_K_M (file size 2.46 GB, estimated 6.3 GB memory). For 12 GB VRAM, Q5_K_M (2.84 GB, 6.7 GB memory) or Q6_K (3.25 GB, 7.1 GB memory) are suitable. Higher quants increase quality but require more memory.

What GPU do you need for Sailor 4B?

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: the best quant that fits is Q3_K_L (2.2 GB file, about 6.0 GB in use).
  • 8 GB VRAM: the best quant that fits is Q6_K (3.2 GB file, about 7.1 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (4.2 GB file, about 8.1 GB in use).

Which Sailor 4B quant should you choose?

Q4_K_M is the usual starting point for Sailor 4B: it balances file size and quality well. Here it is a 2.5 GB file.

The Q8_0 file is 4.2 GB, about 1.7 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 1.6 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

Architectureqwen2
Parameters3.95B
Layers40
Trained context32,768 tokens
KV cache (fp16)400 KB per token

Source

Quantized byRichardErkhov
FamilySailor
Licenseapache-2.0
Downloads (HF)639
Last updated2024-08-19

Frequently asked questions about Sailor 4B

How much VRAM does Sailor 4B need?

About 6.3 GB at Q4_K_M with an 8K context window. That is the 2.5 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 Sailor 4B on an 8 GB GPU?

Yes, with Q6_K: the 3.2 GB file needs about 7.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Sailor 4B GGUF?

It depends on the quant: Q2_K is 1.6 GB, IQ3_XS is 1.8 GB, IQ3_S is 1.9 GB, Q3_K_S is 1.9 GB, IQ3_M is 1.9 GB, Q3_K is 2.0 GB and 16 more.

What context length does Sailor 4B support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 3.4 GB at 8K and 13.4 GB at 32K.

Where can I download Sailor 4B GGUF?

The files are on Hugging Face in the RichardErkhov/sail_-_Sailor-4B-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 Sailor 4B 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 Sailor‑4B on 8 GB VRAM?

Yes, using Q4_K_M which needs about 6.3 GB memory at 8k context. Other quants will exceed 8 GB.

Does Sailor‑4B support vision tasks?

No, the model has no vision capabilities.

What license applies to Sailor‑4B?

The model is released under the apache‑2.0 license.

Similar models

Browse all →

More from RichardErkhov

All by RichardErkhov →

More Sailor models

All Sailor →

Other small models

All small models →