Hugging Face's logo

mradermacher
/
Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF

Transformers
GGUF
English
dag-reasoning
valiant
valiant-labs
qwen
qwen-3
qwen-3-4b
qwen3-4b-thinking-2507
4b
thinking
reasoning
directed-acyclic-graph
graph
logic
analysis
programming
knowledge
root-cause-analysis
economics
business
business-management
finance
law
supply-chain
logistics
software-engineering
cybersecurity
architecture
energy
politics
problem-solving
creative
analytical
expert
rationality
conversational
chat
instruct
imatrix
Model card Files
xet
Community
mradermacher commited on
Commit
372df47
·
verified ·
1 Parent(s): 7916372

auto-patch README.md

Browse files
Files changed (1) hide show
  1. README.md +116 -0
README.md CHANGED
@@ -1,3 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  <!-- ### quantize_version: 2 -->
2
  <!-- ### output_tensor_quantised: 1 -->
3
  <!-- ### convert_type: hf -->
@@ -7,3 +59,67 @@
7
  <!-- ### quants_skip: -->
8
  <!-- ### skip_mmproj: -->
9
  weighted/imatrix quants of https://huggingface.co/sequelbox/Qwen3-4B-Thinking-2507-DAG-Reasoning
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: sequelbox/Qwen3-4B-Thinking-2507-DAG-Reasoning
3
+ datasets:
4
+ - sequelbox/DAG-Reasoning-DeepSeek-R1-0528
5
+ language:
6
+ - en
7
+ library_name: transformers
8
+ license: apache-2.0
9
+ mradermacher:
10
+ readme_rev: 1
11
+ quantized_by: mradermacher
12
+ tags:
13
+ - dag-reasoning
14
+ - valiant
15
+ - valiant-labs
16
+ - qwen
17
+ - qwen-3
18
+ - qwen-3-4b
19
+ - qwen3-4b-thinking-2507
20
+ - 4b
21
+ - thinking
22
+ - reasoning
23
+ - directed-acyclic-graph
24
+ - graph
25
+ - logic
26
+ - analysis
27
+ - programming
28
+ - knowledge
29
+ - root-cause-analysis
30
+ - economics
31
+ - business
32
+ - business-management
33
+ - finance
34
+ - law
35
+ - supply-chain
36
+ - logistics
37
+ - software-engineering
38
+ - cybersecurity
39
+ - architecture
40
+ - energy
41
+ - politics
42
+ - problem-solving
43
+ - creative
44
+ - analytical
45
+ - expert
46
+ - rationality
47
+ - conversational
48
+ - chat
49
+ - instruct
50
+ ---
51
+ ## About
52
+
53
  <!-- ### quantize_version: 2 -->
54
  <!-- ### output_tensor_quantised: 1 -->
55
  <!-- ### convert_type: hf -->
 
59
  <!-- ### quants_skip: -->
60
  <!-- ### skip_mmproj: -->
61
  weighted/imatrix quants of https://huggingface.co/sequelbox/Qwen3-4B-Thinking-2507-DAG-Reasoning
62
+
63
+ <!-- provided-files -->
64
+
65
+ ***For a convenient overview and download list, visit our [model page for this model](https://hf.tst.eu/model#Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF).***
66
+
67
+ static quants are available at https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-GGUF
68
+ ## Usage
69
+
70
+ If you are unsure how to use GGUF files, refer to one of [TheBloke's
71
+ READMEs](https://huggingface.co/TheBloke/KafkaLM-70B-German-V0.1-GGUF) for
72
+ more details, including on how to concatenate multi-part files.
73
+
74
+ ## Provided Quants
75
+
76
+ (sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)
77
+
78
+ | Link | Type | Size/GB | Notes |
79
+ |:-----|:-----|--------:|:------|
80
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.imatrix.gguf) | imatrix | 0.1 | imatrix file (for creating your own qwuants) |
81
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ1_S.gguf) | i1-IQ1_S | 1.2 | for the desperate |
82
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ1_M.gguf) | i1-IQ1_M | 1.2 | mostly desperate |
83
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ2_XXS.gguf) | i1-IQ2_XXS | 1.3 | |
84
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ2_XS.gguf) | i1-IQ2_XS | 1.5 | |
85
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ2_S.gguf) | i1-IQ2_S | 1.5 | |
86
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ2_M.gguf) | i1-IQ2_M | 1.6 | |
87
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q2_K_S.gguf) | i1-Q2_K_S | 1.7 | very low quality |
88
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q2_K.gguf) | i1-Q2_K | 1.8 | IQ3_XXS probably better |
89
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ3_XXS.gguf) | i1-IQ3_XXS | 1.8 | lower quality |
90
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ3_XS.gguf) | i1-IQ3_XS | 1.9 | |
91
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q3_K_S.gguf) | i1-Q3_K_S | 2.0 | IQ3_XS probably better |
92
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ3_S.gguf) | i1-IQ3_S | 2.0 | beats Q3_K* |
93
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ3_M.gguf) | i1-IQ3_M | 2.1 | |
94
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q3_K_M.gguf) | i1-Q3_K_M | 2.2 | IQ3_S probably better |
95
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q3_K_L.gguf) | i1-Q3_K_L | 2.3 | IQ3_M probably better |
96
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ4_XS.gguf) | i1-IQ4_XS | 2.4 | |
97
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q4_0.gguf) | i1-Q4_0 | 2.5 | fast, low quality |
98
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-IQ4_NL.gguf) | i1-IQ4_NL | 2.5 | prefer IQ4_XS |
99
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q4_K_S.gguf) | i1-Q4_K_S | 2.5 | optimal size/speed/quality |
100
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q4_K_M.gguf) | i1-Q4_K_M | 2.6 | fast, recommended |
101
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q4_1.gguf) | i1-Q4_1 | 2.7 | |
102
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q5_K_S.gguf) | i1-Q5_K_S | 2.9 | |
103
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q5_K_M.gguf) | i1-Q5_K_M | 3.0 | |
104
+ | [GGUF](https://huggingface.co/mradermacher/Qwen3-4B-Thinking-2507-DAG-Reasoning-i1-GGUF/resolve/main/Qwen3-4B-Thinking-2507-DAG-Reasoning.i1-Q6_K.gguf) | i1-Q6_K | 3.4 | practically like static Q6_K |
105
+
106
+ Here is a handy graph by ikawrakow comparing some lower-quality quant
107
+ types (lower is better):
108
+
109
+ ![image.png](https://www.nethype.de/huggingface_embed/quantpplgraph.png)
110
+
111
+ And here are Artefact2's thoughts on the matter:
112
+ https://gist.github.com/Artefact2/b5f810600771265fc1e39442288e8ec9
113
+
114
+ ## FAQ / Model Request
115
+
116
+ See https://huggingface.co/mradermacher/model_requests for some answers to
117
+ questions you might have and/or if you want some other model quantized.
118
+
119
+ ## Thanks
120
+
121
+ I thank my company, [nethype GmbH](https://www.nethype.de/), for letting
122
+ me use its servers and providing upgrades to my workstation to enable
123
+ this work in my free time. Additional thanks to [@nicoboss](https://huggingface.co/nicoboss) for giving me access to his private supercomputer, enabling me to provide many more imatrix quants, at much higher quality, than I would otherwise be able to.
124
+
125
+ <!-- end -->