codegeex4-all-9b needs about 7.1 GB of memory at Q4_K_M (a 6.3 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 23 quants, from 3.6 GB (IQ2_XS) to 37.6 GB (F32). Longer context needs more: at 32K tokens the KV cache alone adds about 1.3 GB.
9.4B Codegeex model for code generation
codegeex4-all-9b is a 9.4B parameter chatglm model optimized for code generation. It supports 131072 context tokens and is available in multiple GGUF quantizations. The model is released under an other license.
Good at
- Code completion
- Bug detection
- API generation
Limits
- No vision support
- Large context requires high memory
- License not specified
Best for
- Developers writing code
- Automated code review
- Learning new APIs
Who it’s for. Engineers and developers who need a large language model focused on code tasks. The model is suitable for local deployment using llama.cpp.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_XS | 3.6 GB | 4.4 GB | Fits | Get file → | |
| IQ2_S | 3.8 GB | 4.6 GB | Fits | Get file → | |
| IQ2_M | 3.9 GB | 4.8 GB | Fits | Get file → | |
| Q2_K | 4.0 GB | 4.8 GB | Fits | Get file → | |
| IQ3_XXS | 4.3 GB | 5.1 GB | Fits | Get file → | |
| IQ3_XS | 4.4 GB | 5.3 GB | Fits | Get file → | |
| Q3_K_S | 4.6 GB | 5.4 GB | Fits | Get file → | |
| Q2_K_L | 4.6 GB | 5.4 GB | Fits | Get file → | |
| IQ3_M | 4.8 GB | 5.6 GB | Fits | Get file → | |
| Q3_K_M | 5.1 GB | 5.9 GB | Fits | Get file → | |
| IQ4_XS | 5.3 GB | 6.1 GB | Fits | Get file → | |
| Q3_K_L | 5.3 GB | 6.1 GB | Fits | Get file → | |
| Q4_K_S | 5.8 GB | 6.6 GB | Fits | Get file → | |
| Q3_K_XL | 5.8 GB | 6.7 GB | Fits | Get file → | |
| Q4_K_M | 6.3 GB | 7.1 GB | Fits | Get file → | |
| Q5_K_S | 6.7 GB | 7.5 GB | Tight | Get file → | |
| Q4_K_L | 6.7 GB | 7.5 GB | Tight | Get file → | |
| Q5_K_M | 7.1 GB | 8.0 GB | Tight | Get file → | |
| Q5_K_L | 7.5 GB | 8.4 GB | Too big | Get file → | |
| Q6_K | 8.3 GB | 9.1 GB | Too big | Get file → | |
| Q6_K_L | 8.6 GB | 9.4 GB | Too big | Get file → | |
| Q8_0 | 10.0 GB | 10.8 GB | Too big | Get file → | |
| F32 | 37.6 GB | 38.4 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_S (file size 5.75 GB, memory 6.6 GB). For 12 GB VRAM, Q5_K_S (6.69 GB, 7.5 GB) or Q4_K_M (6.25 GB, 7.1 GB). For 16 GB VRAM, Q5_K_M (7.14 GB, 8 GB) or Q5_K_L (7.53 GB, 8.4 GB).
What GPU do you need for codegeex4-all-9b?
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 (5.1 GB file, about 5.9 GB in use).
- 8 GB VRAM: the best quant that fits is Q5_K_M (7.1 GB file, about 8.0 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (10.0 GB file, about 10.8 GB in use).
- 16 GB VRAM: the best quant that fits is Q8_0 (10.0 GB file, about 10.8 GB in use).
- 24 GB VRAM: the best quant that fits is Q8_0 (10.0 GB file, about 10.8 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (10.0 GB file, about 10.8 GB in use).
- 48 GB VRAM: the best quant that fits is F32 (37.6 GB file, about 38.4 GB in use).
Which codegeex4-all-9b quant should you choose?
Q4_K_M is the usual starting point for codegeex4-all-9b: it balances file size and quality well. Here it is a 6.3 GB file.
The Q8_0 file is 10.0 GB, about 1.6 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 IQ2_XS at 3.6 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
Source
Frequently asked questions about codegeex4-all-9b
How much VRAM does codegeex4-all-9b need?
About 7.1 GB at Q4_K_M with an 8K context window. That is the 6.3 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 codegeex4-all-9b on an 8 GB GPU?
Yes, with Q5_K_M: the 7.1 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 codegeex4-all-9b GGUF?
It depends on the quant: IQ2_XS is 3.6 GB, IQ2_S is 3.8 GB, IQ2_M is 3.9 GB, Q2_K is 4.0 GB, IQ3_XXS is 4.3 GB, IQ3_XS is 4.4 GB and 17 more.
What context length does codegeex4-all-9b support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.3 GB at 8K and 1.3 GB at 32K.
Where can I download codegeex4-all-9b GGUF?
The files are on Hugging Face in the bartowski/codegeex4-all-9b-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 codegeex4-all-9b use?
Hugging Face lists the license as other. 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 codegeex4-all-9b on 8 GB VRAM?
Yes, a K-quant such as Q4_K_S fits within 8 GB VRAM with an estimated 6.6 GB usage.
What is the smallest quant available?
IQ2_XS is the smallest at 3.61 GB file size and 4.4 GB estimated memory.
Does the model support vision?
No, the model has no vision capabilities.