KO

KoSOLAR 10.9B v1.0

KosolarGGUF by mradermacherUpdated 2024-05-06
Download on Hugging Face
Parameters10.9B
Context4K
Layers48
Architecturellama
Smallest quant4.7 GB
Quick answer

KoSOLAR 10.9B v1.0 needs about 9.3 GB of memory at Q4_K_M (a 7.2 GB file) with an 8K context window, so it fits on a 12 GB GPU. This repository has 16 quants, from 4.7 GB (Q2_K) to 12.1 GB (Q8_0).

KoSOLAR 10.9B v1.0 – a 10.9B parameter Llama‑style model

KoSOLAR 10.9B v1.0 is a 10.9B parameter Llama architecture model trained with a 4096 token context. It is released under the Apache‑2.0 license and is available in multiple GGUF quantizations. The model is suitable for general-purpose language tasks.

Good at

  • General text generation
  • Contextual understanding up to 4096 tokens
  • Multilingual support

Limits

  • No vision capabilities
  • Requires at least 6.8 GB VRAM for Q2_K
  • License is Apache‑2.0

Best for

  • Chatbots with moderate memory
  • Fine‑tuning on custom datasets
  • Research prototyping

Who it’s for. Developers and researchers needing a large Llama‑style model with flexible quantization options. The model is ideal for those who can manage the memory requirements and want an Apache‑2.0 licensed asset.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K4.7 GB 6.0 GB
Fits Get file →
IQ3_XS5.2 GB 6.5 GB
Fits Get file →
Q3_K_S5.4 GB 6.7 GB
Fits Get file →
IQ3_S5.4 GB 6.7 GB
Fits Get file →
IQ3_M5.6 GB 6.9 GB
Fits Get file →
Q3_K_M5.9 GB 7.2 GB
Tight Get file →
Q3_K_L6.4 GB 7.7 GB
Tight Get file →
IQ4_XS6.5 GB 7.8 GB
Tight Get file →
Q4_06.8 GB 8.1 GB
Too big Get file →
Q4_K_S6.8 GB 8.1 GB
Too big Get file →
IQ4_NL6.9 GB 8.2 GB
Too big Get file →
Q4_K_M7.2 GB 8.5 GB
Too big Get file →
Q5_K_S8.1 GB 9.4 GB
Too big Get file →
Q5_K_M8.3 GB 9.6 GB
Too big Get file →
Q6_K9.5 GB 10.8 GB
Too big Get file →
Q8_012.1 GB 13.4 GB
Too big Get file →

For 8 GB VRAM, use Q2_K (file size 4.7 GB, estimated 6.8 GB memory). If 10 GB VRAM is available, Q3_K_S (5.37 GB, 7.5 GB memory) or IQ3_S (5.4 GB, 7.5 GB memory) are good choices. For 14 GB VRAM, Q8_0 (12.12 GB, 14.2 GB memory) offers the highest quality.

What GPU do you need for KoSOLAR 10.9B v1.0?

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 Q3_K_M (5.9 GB file, about 8.0 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (9.5 GB file, about 11.6 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (12.1 GB file, about 14.2 GB in use).

Which KoSOLAR 10.9B v1.0 quant should you choose?

Q4_K_M is the usual starting point for KoSOLAR 10.9B v1.0: it balances file size and quality well. Here it is a 7.2 GB file.

The Q8_0 file is 12.1 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 4.7 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
Parameters10.9B
Layers48
Trained context4,096 tokens
KV cache (fp16)192 KB per token

Source

Quantized bymradermacher
FamilyKosolar
Licenseapache-2.0
Downloads (HF)1,620
Last updated2024-05-06

Frequently asked questions about KoSOLAR 10.9B v1.0

How much VRAM does KoSOLAR 10.9B v1.0 need?

About 9.3 GB at Q4_K_M with an 8K context window. That is the 7.2 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 KoSOLAR 10.9B v1.0 on an 8 GB GPU?

Yes, with Q3_K_M: the 5.9 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of KoSOLAR 10.9B v1.0 GGUF?

It depends on the quant: Q2_K is 4.7 GB, IQ3_XS is 5.2 GB, Q3_K_S is 5.4 GB, IQ3_S is 5.4 GB, IQ3_M is 5.6 GB, Q3_K_M is 5.9 GB and 10 more.

What context length does KoSOLAR 10.9B v1.0 support?

The GGUF header lists a trained context length of 4,096 tokens (4K). The KV cache grows with the context you use: about 1.6 GB at 8K and 0.8 GB at 4K.

Where can I download KoSOLAR 10.9B v1.0 GGUF?

The files are on Hugging Face in the mradermacher/KoSOLAR-10.9B-v1.0-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 KoSOLAR 10.9B v1.0 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 KoSOLAR 10.9B v1.0 on 8 GB VRAM?

Yes, using the Q2_K quantization which requires approximately 6.8 GB VRAM at 8k context.

Does KoSOLAR support vision tasks?

No, the model does not include vision capabilities.

Other GGUFs of KoSOLAR-10.9B-v1.0

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Kosolar models

All Kosolar →

Other medium models

All medium models →