CA

CantoneseLLM v2.0 8B Thinking Chat Vector Merged

CantonesellmGGUF by mradermacherUpdated 2026-09-09
Download on Hugging Face
Parameters8.19B
Context40K
Layers36
Architectureqwen3
Smallest quant3.3 GB
Quick answer

CantoneseLLM v2.0 8B Thinking Chat Vector Merged needs about 6.7 GB of memory at Q4_K_M (a 5.0 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 12 quants, from 3.3 GB (Q2_K) to 16.4 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.8 GB.

CantoneseLLM v2.0 8B – an 8B Qwen3 model with Cantonese knowledge and instruction following

CantoneseLLM v2.0 8B Thinking Chat Vector Merged is a dense 8.19B parameter model built from Qwen3-8B-Base. It was continually pre‑trained on 784M Cantonese and Hong Kong text and merged with a chat vector from Qwen3-8B. The model follows instructions and can reason, but its chain‑of‑thought is in Simplified Chinese.

Good at

  • Cantonese knowledge
  • Instruction following
  • Reasoning in Simplified Chinese

Limits

  • No Cantonese chain‑of‑thought
  • No vision support
  • Apache‑2.0 license

Best for

  • Chatting in Cantonese
  • Instruction following tasks
  • Cantonese text generation

Who it’s for. Developers needing a Cantonese‑knowledgeable 8B instruct model without requiring Cantonese reasoning. It is suitable for research reproductions of the paper’s ablations or for applications where Simplified Chinese reasoning is acceptable.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q2_K3.3 GB 5.0 GB
Fits Get file →
Q3_K_S3.8 GB 5.5 GB
Fits Get file →
Q3_K_M4.1 GB 5.8 GB
Fits Get file →
Q3_K_L4.4 GB 6.1 GB
Fits Get file →
IQ4_XS4.6 GB 6.3 GB
Fits Get file →
Q4_K_S4.8 GB 6.5 GB
Fits Get file →
Q4_K_M5.0 GB 6.7 GB
Fits Get file →
Q5_K_S5.7 GB 7.4 GB
Tight Get file →
Q5_K_M5.9 GB 7.6 GB
Tight Get file →
Q6_K6.7 GB 8.4 GB
Too big Get file →
Q8_08.7 GB 10.4 GB
Too big Get file →
F1616.4 GB 18.1 GB
Too big Get file →

For 8 GB VRAM, use Q2_K (3.28 GB) or Q3_K_S (3.77 GB). For 12 GB VRAM, Q4_K_M (5.03 GB) or Q5_K_M (5.85 GB) are suitable. For 16 GB VRAM, Q6_K (6.73 GB) or Q8_0 (8.71 GB) provide higher quality. Memory estimates are approximate.

What GPU do you need for CantoneseLLM v2.0 8B Thinking Chat Vector Merged?

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_M (4.1 GB file, about 5.8 GB in use).
  • 8 GB VRAM: the best quant that fits is Q5_K_M (5.9 GB file, about 7.6 GB in use).
  • 12 GB VRAM: the best quant that fits is Q8_0 (8.7 GB file, about 10.4 GB in use).
  • 16 GB VRAM: the best quant that fits is Q8_0 (8.7 GB file, about 10.4 GB in use).
  • 24 GB VRAM: the best quant that fits is F16 (16.4 GB file, about 18.1 GB in use).

Which CantoneseLLM v2.0 8B Thinking Chat Vector Merged quant should you choose?

Q4_K_M is the usual starting point for CantoneseLLM v2.0 8B Thinking Chat Vector Merged: it balances file size and quality well. Here it is a 5.0 GB file.

The Q8_0 file is 8.7 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 3.3 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

Architectureqwen3
Parameters8.19B
Layers36
Trained context40,960 tokens
KV cache (fp16)144 KB per token

Frequently asked questions about CantoneseLLM v2.0 8B Thinking Chat Vector Merged

How much VRAM does CantoneseLLM v2.0 8B Thinking Chat Vector Merged need?

About 6.7 GB at Q4_K_M with an 8K context window. That is the 5.0 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 CantoneseLLM v2.0 8B Thinking Chat Vector Merged on an 8 GB GPU?

Yes, with Q5_K_M: the 5.9 GB file needs about 7.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of CantoneseLLM v2.0 8B Thinking Chat Vector Merged GGUF?

It depends on the quant: Q2_K is 3.3 GB, Q3_K_S is 3.8 GB, Q3_K_M is 4.1 GB, Q3_K_L is 4.4 GB, IQ4_XS is 4.6 GB, Q4_K_S is 4.8 GB and 6 more.

What context length does CantoneseLLM v2.0 8B Thinking Chat Vector Merged support?

The GGUF header lists a trained context length of 40,960 tokens (40K). The KV cache grows with the context you use: about 1.2 GB at 8K and 4.8 GB at 32K.

Where can I download CantoneseLLM v2.0 8B Thinking Chat Vector Merged GGUF?

The files are on Hugging Face in the mradermacher/CantoneseLLM-v2.0-8B-Thinking-Chat-Vector-Merged-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 CantoneseLLM v2.0 8B Thinking Chat Vector Merged 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 this model on 8 GB VRAM?

Yes, the Q2_K and Q3_K_S quantizations fit within 8 GB VRAM, with estimated memory usage of 5 GB and 5.5 GB respectively at 8K context.

Does this model support vision tasks?

No, the model does not have vision capabilities.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Other GGUFs of CantoneseLLM-v2.0-8B-Thinking-Chat-Vector-Merged

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Cantonesellm models

All Cantonesellm →

Other small models

All small models →