[gMASK]<sop>
<|system|>
你是一个专业的深度研究助手，通过提供的工具与模拟浏览器交互，来帮助用户完成深度信息调研和报告撰写任务。今年是 2025 年。

<核心要求>
- 首先分解用户请求，得到包含多个子要求的列表
- 制定初始研究计划
- 进行多轮迭代搜索和页面浏览（at least 10 function calls）：
    * 根据已获得的信息调整研究计划和关键词
    * 打开页面阅读，从发现的内容中识别新的关键概念/名词
    * 从搜索结果中提取新的关键词继续搜索
    * 访问并仔细阅读相关页面，识别新的关键概念/名词

<重要配置>
- 采用语言
    * 搜索关键词：英文
    * 思考：英文

<可调用的工具列表>
[{"name": "search", "description": "Execute a search query and return search results. Use this function when you need to find information about a specific topic.", "parameters": {"type": "object", "properties": {"query": {"type": "string", "description": "Search query string, use English words unless it is a proper name in Chinese"}}, "required": ["query"], "additionalProperties": false}}, {"name": "click", "description": "Click a link in the search results and navigate to the corresponding page. Use this function when you need to view detailed content of a specific search result.", "parameters": {"type": "object", "properties": {"link_id": {"type": "integer", "description": "The link ID to click (from the sequence number in search results)"}}, "required": ["link_id"], "additionalProperties": false}}, {"name": "open", "description": "Open a specific website. Get content from any website with its URL.", "parameters": {"type": "object", "properties": {"url": {"type": "string", "description": "The target website URL or domain"}}, "required": ["url"], "additionalProperties": false}}, {"name": "finish", "description": "Finish the task. Use this function when you have found the information you need.", "parameters": {"type": "object", "properties": {}, "additionalProperties": false}}]<|user|>
{prompt}<|assistant|>

What's new:

Fix tokenizer

Download a file (not the whole branch) from below:

Filename	Quant type	File Size	Split	Description
GLM-Z1-Rumination-32B-0414-bf16.gguf	bf16	66.30GB	true	Full BF16 weights.
GLM-Z1-Rumination-32B-0414-Q8_0.gguf	Q8_0	35.23GB	false	Extremely high quality, generally unneeded but max available quant.
GLM-Z1-Rumination-32B-0414-Q6_K_L.gguf	Q6_K_L	27.65GB	false	Uses Q8_0 for embed and output weights. Very high quality, near perfect, recommended.
GLM-Z1-Rumination-32B-0414-Q6_K.gguf	Q6_K	27.20GB	false	Very high quality, near perfect, recommended.
GLM-Z1-Rumination-32B-0414-Q5_K_L.gguf	Q5_K_L	24.09GB	false	Uses Q8_0 for embed and output weights. High quality, recommended.
GLM-Z1-Rumination-32B-0414-Q5_K_M.gguf	Q5_K_M	23.51GB	false	High quality, recommended.
GLM-Z1-Rumination-32B-0414-Q5_K_S.gguf	Q5_K_S	22.92GB	false	High quality, recommended.
GLM-Z1-Rumination-32B-0414-Q4_1.gguf	Q4_1	20.91GB	false	Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon.
GLM-Z1-Rumination-32B-0414-Q4_K_L.gguf	Q4_K_L	20.73GB	false	Uses Q8_0 for embed and output weights. Good quality, recommended.
GLM-Z1-Rumination-32B-0414-Q4_K_M.gguf	Q4_K_M	20.04GB	false	Good quality, default size for most use cases, recommended.
GLM-Z1-Rumination-32B-0414-Q4_K_S.gguf	Q4_K_S	19.02GB	false	Slightly lower quality with more space savings, recommended.
GLM-Z1-Rumination-32B-0414-Q4_0.gguf	Q4_0	18.96GB	false	Legacy format, offers online repacking for ARM and AVX CPU inference.
GLM-Z1-Rumination-32B-0414-IQ4_NL.gguf	IQ4_NL	18.94GB	false	Similar to IQ4_XS, but slightly larger. Offers online repacking for ARM CPU inference.
GLM-Z1-Rumination-32B-0414-Q3_K_XL.gguf	Q3_K_XL	18.35GB	false	Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability.
GLM-Z1-Rumination-32B-0414-IQ4_XS.gguf	IQ4_XS	17.95GB	false	Decent quality, smaller than Q4_K_S with similar performance, recommended.
GLM-Z1-Rumination-32B-0414-Q3_K_L.gguf	Q3_K_L	17.54GB	false	Lower quality but usable, good for low RAM availability.
GLM-Z1-Rumination-32B-0414-Q3_K_M.gguf	Q3_K_M	16.18GB	false	Low quality.
GLM-Z1-Rumination-32B-0414-IQ3_M.gguf	IQ3_M	15.11GB	false	Medium-low quality, new method with decent performance comparable to Q3_K_M.
GLM-Z1-Rumination-32B-0414-Q3_K_S.gguf	Q3_K_S	14.62GB	false	Low quality, not recommended.
GLM-Z1-Rumination-32B-0414-IQ3_XS.gguf	IQ3_XS	13.93GB	false	Lower quality, new method with decent performance, slightly better than Q3_K_S.
GLM-Z1-Rumination-32B-0414-Q2_K_L.gguf	Q2_K_L	13.46GB	false	Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable.
GLM-Z1-Rumination-32B-0414-IQ3_XXS.gguf	IQ3_XXS	13.04GB	false	Lower quality, new method with decent performance, comparable to Q3 quants.
GLM-Z1-Rumination-32B-0414-Q2_K.gguf	Q2_K	12.55GB	false	Very low quality but surprisingly usable.
GLM-Z1-Rumination-32B-0414-IQ2_M.gguf	IQ2_M	11.53GB	false	Relatively low quality, uses SOTA techniques to be surprisingly usable.
GLM-Z1-Rumination-32B-0414-IQ2_S.gguf	IQ2_S	10.67GB	false	Low quality, uses SOTA techniques to be usable.
GLM-Z1-Rumination-32B-0414-IQ2_XS.gguf	IQ2_XS	10.15GB	false	Low quality, uses SOTA techniques to be usable.

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

Downloading using huggingface-cli

Click to view download instructions

First, make sure you have hugginface-cli installed:

pip install -U "huggingface_hub[cli]"

Then, you can target the specific file you want:

huggingface-cli download bartowski/THUDM_GLM-Z1-Rumination-32B-0414-GGUF --include "THUDM_GLM-Z1-Rumination-32B-0414-Q4_K_M.gguf" --local-dir ./

If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

huggingface-cli download bartowski/THUDM_GLM-Z1-Rumination-32B-0414-GGUF --include "THUDM_GLM-Z1-Rumination-32B-0414-Q8_0/*" --local-dir ./

You can either specify a new local-dir (THUDM_GLM-Z1-Rumination-32B-0414-Q8_0) or download them all in place (./)

ARM/AVX information

Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.

Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.

As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.

Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.

Click to view Q4_0_X_X information (deprecated

I'm keeping this section to show the potential theoretical uplift in performance from using the Q4_0 with online repacking.

Click to view benchmarks on an AVX2 system (EPYC7702)

model	size	params	backend	threads	test	t/s	% (vs Q4_0)
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp512	204.03 ± 1.03	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp1024	282.92 ± 0.19	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	pp2048	259.49 ± 0.44	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg128	39.12 ± 0.27	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg256	39.31 ± 0.69	100%
qwen2 3B Q4_0	1.70 GiB	3.09 B	CPU	64	tg512	40.52 ± 0.03	100%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp512	301.02 ± 1.74	147%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp1024	287.23 ± 0.20	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	pp2048	262.77 ± 1.81	101%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg128	18.80 ± 0.99	48%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg256	24.46 ± 3.04	83%
qwen2 3B Q4_K_M	1.79 GiB	3.09 B	CPU	64	tg512	36.32 ± 3.59	90%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp512	271.71 ± 3.53	133%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp1024	279.86 ± 45.63	100%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	pp2048	320.77 ± 5.00	124%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg128	43.51 ± 0.05	111%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg256	43.35 ± 0.09	110%
qwen2 3B Q4_0_8_8	1.69 GiB	3.09 B	CPU	64	tg512	42.60 ± 0.31	105%

Q4_0_8_8 offers a nice bump to prompt processing and a small bump to text generation

Which file should I choose?

Click here for details

A great write up with charts showing various performances is provided by Artefact2 here

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.

If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.

Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.

If you don't want to think too much, grab one of the K-quants. These are in format 'QX_K_X', like Q5_K_M.

If you want to get more into the weeds, you can check out this extremely useful feature chart:

llama.cpp feature matrix

But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQX_X, like IQ3_M. These are newer and offer better performance for their size.

These I-quants can also be used on CPU, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.

Credits

Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset.

Thank you ZeroWw for the inspiration to experiment with embed/output.

Thank you to LM Studio for sponsoring my work.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski