Hugging Face's logo

LGAI-EXAONE
/
K-EXAONE-2.0-750B-A37B-DSpark

Text Generation
Transformers
Safetensors
exaone_moe
lg-ai
exaone
k-exaone
Mixture of Experts
conversational
Model card Files
xet
 Community
LG-AI-EXAONE commited on
Commit
5e72b87
·
0 Parent(s):

Initial commit

Browse files
Files changed (31) hide show
  1. .gitattributes +40 -0
  2. LICENSE +15 -0
  3. NOTICE +14 -0
  4. README.md +728 -0
  5. assets/K-EXAONE-2.0-Technical-Report.pdf +3 -0
  6. assets/K-EXAONE_logo_gray.png +3 -0
  7. assets/k-ai-logo-no-bg.png +3 -0
  8. assets/main_figure.png +3 -0
  9. chat_template.jinja +194 -0
  10. config.json +391 -0
  11. generation_config.json +9 -0
  12. model-00001-of-00016.safetensors +3 -0
  13. model-00002-of-00016.safetensors +3 -0
  14. model-00003-of-00016.safetensors +3 -0
  15. model-00004-of-00016.safetensors +3 -0
  16. model-00005-of-00016.safetensors +3 -0
  17. model-00006-of-00016.safetensors +3 -0
  18. model-00007-of-00016.safetensors +3 -0
  19. model-00008-of-00016.safetensors +3 -0
  20. model-00009-of-00016.safetensors +3 -0
  21. model-00010-of-00016.safetensors +3 -0
  22. model-00011-of-00016.safetensors +3 -0
  23. model-00012-of-00016.safetensors +3 -0
  24. model-00013-of-00016.safetensors +3 -0
  25. model-00014-of-00016.safetensors +3 -0
  26. model-00015-of-00016.safetensors +3 -0
  27. model-00016-of-00016.safetensors +3 -0
  28. model-dspark.safetensors +3 -0
  29. model.safetensors.index.json +0 -0
  30. tokenizer.json +3 -0
  31. tokenizer_config.json +19 -0
.gitattributes ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ assets/K-EXAONE_logo_gray.png filter=lfs diff=lfs merge=lfs -text
37
+ assets/k-ai-logo-no-bg.png filter=lfs diff=lfs merge=lfs -text
38
+ assets/main_figure.png filter=lfs diff=lfs merge=lfs -text
39
+ tokenizer.json filter=lfs diff=lfs merge=lfs -text
40
+ assets/K-EXAONE-2.0-Technical-Report.pdf filter=lfs diff=lfs merge=lfs -text
LICENSE ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Apache 2.0
2
+
3
+ Copyright © 2026 LG AI Research
4
+
5
+ Licensed under the Apache License, Version 2.0 (the "License");
6
+ you may not use this file except in compliance with the License.
7
+ You may obtain a copy of the License at
8
+
9
+ http://www.apache.org/licenses/LICENSE-2.0
10
+
11
+ Unless required by applicable law or agreed to in writing, software
12
+ distributed under the License is distributed on an "AS IS" BASIS,
13
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ See the License for the specific language governing permissions and
15
+ limitations under the License.
NOTICE ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ K-EXAONE 2.0
2
+ Copyright © 2026 LG AI Research
3
+
4
+ Licensed under the Apache License, Version 2.0 (the "License");
5
+ you may not use this file except in compliance with the License.
6
+ You may obtain a copy of the License at
7
+
8
+ http://www.apache.org/licenses/LICENSE-2.0
9
+
10
+ Unless required by applicable law or agreed to in writing, software
11
+ distributed under the License is distributed on an "AS IS" BASIS,
12
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ See the License for the specific language governing permissions and
14
+ limitations under the License.
README.md ADDED
@@ -0,0 +1,728 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+
3
+ language:
4
+ - en
5
+ - ko
6
+ - es
7
+ - de
8
+ - ja
9
+ - vi
10
+ - fr
11
+ - it
12
+ - pl
13
+ - pt
14
+ license: apache-2.0
15
+ license_link: LICENSE
16
+ library_name: transformers
17
+ pipeline_tag: text-generation
18
+ tags:
19
+ - lg-ai
20
+ - exaone
21
+ - k-exaone
22
+ - moe
23
+ ---
24
+
25
+ # K-EXAONE-2.0-750B-A37B-DSpark
26
+ <br>
27
+ <br>
28
+ <p align="center">
29
+ <img src="assets/K-EXAONE_logo_gray.png" width="400">
30
+ <br>
31
+ <br>
32
+ <br>
33
+
34
+ <div align="center">
35
+ <a href="https://huggingface.co/collections/LGAI-EXAONE/k-exaone-20" style="text-decoration: none;">
36
+ <img src="https://img.shields.io/badge/🤗-HuggingFace-FC926C?style=for-the-badge" alt="HuggingFace">
37
+ </a>
38
+ <a href="#" style="text-decoration: none;">
39
+ <img src="https://img.shields.io/badge/📝-Blog-E343BD?style=for-the-badge" alt="Blog">
40
+ </a>
41
+ <a href="https://huggingface.co/LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark/blob/main/assets/K-EXAONE-2.0-Technical-Report.pdf" style="text-decoration: none;">
42
+ <img src="https://img.shields.io/badge/📑-Technical_Report-684CF4?style=for-the-badge" alt="Technical Report">
43
+ </a>
44
+ <a href="https://github.com/LG-AI-EXAONE/K-EXAONE-2.0" style="text-decoration: none;">
45
+ <img src="https://img.shields.io/badge/🖥️-GitHub-2B3137?style=for-the-badge" alt="GitHub">
46
+ </a>
47
+ <!-- <a href="#" style="text-decoration: none;">
48
+ <img src="https://img.shields.io/badge/✈️_API-Try_on_FriendliAI-2649BC?style=for-the-badge" alt="FriendliAI">
49
+ </a> -->
50
+ </div>
51
+
52
+
53
+
54
+ ## Introduction
55
+
56
+ We introduce **K-EXAONE 2.0**, a frontier-scale multilingual language model developed by LG AI Research. K-EXAONE 2.0 was scaled to more than three times the size of its predecessor through upcycling, followed by continual pretraining, difficulty-focused mid-training, and post-training.
57
+ K-EXAONE 2.0 is broadly competitive with leading open-weight models, demonstrating substantial improvements over its predecessor and achieving particularly strong results in long-context retrieval and safety.
58
+
59
+ #### Highlights
60
+
61
+ - **Frontier-Class Scale**
62
+ To build a large-scale foundation model with frontier-level intelligence, we upcycled the [K-EXAONE](https://huggingface.co/collections/LGAI-EXAONE/k-exaone) model by expanding both its depth and width, resulting in a more favorable scaling curve. During this process, we found that clamping after two SwiGLU branches effectively mitigates the exploding activations in deeper layers, improving both training and inference stability.
63
+
64
+ - **Advanced Reasoning & Agentic Intelligence**
65
+ In response to the growth of agentic AI, we focused on expanding the model's capabilities in reasoning, agentic workflows, and long-context management. Through careful calibration of the training data and recipes, K-EXAONE 2.0 achieves consistent improvements in agentic coding and long-context understanding, with strong performance on long-context retrieval and safety.
66
+
67
+ - **Production-Ready Inference**
68
+ We support two speculative decoding methods to accelerate inference: MTP (Multi-Token Prediction) and DSpark. Both methods can speed up model generation by approximately 3–5×, reducing latency for long-horizon workloads such as agentic tasks.
69
+
70
+ - **Multilinguality & Openness**
71
+ We expanded multilingual coverage from six to ten languages: Korean, English, Spanish, German, Japanese, Vietnamese, French, Italian, Polish, and Portuguese. We also release K-EXAONE 2.0 under the [Apache license 2.0](https://huggingface.co/LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark/blob/main/LICENSE) so that the broader AI ecosystem can inspect, deploy, and build upon it.
72
+
73
+ ![](assets/main_figure.png)
74
+
75
+ ## Model Configurations
76
+
77
+ <div>
78
+ <table>
79
+ <tr>
80
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Parameters</strong></td>
81
+ <td align="center" style="vertical-align: middle; text-align: center">750B</td>
82
+ </tr>
83
+ <tr>
84
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Active Parameters</strong></td>
85
+ <td align="center" style="vertical-align: middle; text-align: center">37B</td>
86
+ </tr>
87
+ <tr>
88
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Hidden Dimension</strong></td>
89
+ <td align="center" style="vertical-align: middle; text-align: center">6,144</td>
90
+ </tr>
91
+ <tr>
92
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Intermediate Size</strong></td>
93
+ <td align="center" style="vertical-align: middle; text-align: center">18,432</td>
94
+ </tr>
95
+ <tr>
96
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Layers</strong></td>
97
+ <td align="center" style="vertical-align: middle; text-align: center">78 (2 heading Dense + 76 Sparse) Main layers<br>
98
+ 5 DSpark layers<br>
99
+ </td>
100
+ </tr>
101
+ <tr>
102
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Attention</strong></td>
103
+ <td align="center" style="vertical-align: middle; text-align: center">1 x Global (NoPE)<br>
104
+ 1 x 4096 SWA<br>
105
+ 19 x [3 x 128 SWA + 1 x Global] Blocks
106
+ </td>
107
+ </tr>
108
+ <tr>
109
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Attention Heads</strong></td>
110
+ <td align="center" style="vertical-align: middle; text-align: center">64 Q-heads / 8 KV-heads</td>
111
+ </tr>
112
+ <tr>
113
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Head Dimension</strong></td>
114
+ <td align="center" style="vertical-align: middle; text-align: center">128</td>
115
+ </tr>
116
+ <tr>
117
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Number of Experts</strong></td>
118
+ <td align="center" style="vertical-align: middle; text-align: center">1 Shared Expert<br>
119
+ 256 Total Experts<br>
120
+ 8 Activated Experts
121
+ </td>
122
+ </tr>
123
+ <tr>
124
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Expert Dimension</strong></td>
125
+ <td align="center" style="vertical-align: middle; text-align: center">2,048</td>
126
+ </tr>
127
+ <tr>
128
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Vocab Size</strong></td>
129
+ <td align="center" style="vertical-align: middle; text-align: center">153,600</td>
130
+ </tr>
131
+ <tr>
132
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Context Length</strong></td>
133
+ <td align="center" style="vertical-align: middle; text-align: center">262,144</td>
134
+ </tr>
135
+ <tr>
136
+ <td align="center" style="vertical-align: middle; text-align: center"><strong>Knowledge Cutoff</strong></td>
137
+ <td align="center" style="vertical-align: middle; text-align: center">2025 2Q</td>
138
+ </tr>
139
+
140
+ </table>
141
+ </div>
142
+
143
+ ## Evaluation Results
144
+
145
+ The following table shows the benchmark results for the K-EXAONE 2.0 BF16 model.
146
+ Detailed evaluation results and configurations can be found in our [technical report](https://huggingface.co/LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark/blob/main/assets/K-EXAONE-2.0-Technical-Report.pdf).
147
+
148
+ <table>
149
+ <tr>
150
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;"> </th>
151
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;">K-EXAONE 2.0</th>
152
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;">K-EXAONE</th>
153
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;">Qwen3.5</th>
154
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;">GLM-5.1</th>
155
+ <th align="center" style="vertical-align: middle; text-align: center; background: rgba(128,128,128,0.1); text-align: center;">DSV4 Pro&nbsp;(max)</th>
156
+ </tr>
157
+ <tr>
158
+ <td>Architecture</td>
159
+ <td align="center" style="vertical-align: middle; text-align: center; ">MoE</td>
160
+ <td align="center" style="vertical-align: middle; text-align: center; ">MoE</td>
161
+ <td align="center" style="vertical-align: middle; text-align: center; ">MoE</td>
162
+ <td align="center" style="vertical-align: middle; text-align: center; ">MoE</td>
163
+ <td align="center" style="vertical-align: middle; text-align: center; ">MoE</td>
164
+ </tr>
165
+ <tr>
166
+ <td>Total Params</td>
167
+ <td align="center" style="vertical-align: middle; text-align: center; ">750B</td>
168
+ <td align="center" style="vertical-align: middle; text-align: center; ">236B</td>
169
+ <td align="center" style="vertical-align: middle; text-align: center; ">397B</td>
170
+ <td align="center" style="vertical-align: middle; text-align: center; ">754B</td>
171
+ <td align="center" style="vertical-align: middle; text-align: center; ">1.6T</td>
172
+ </tr>
173
+ <tr>
174
+ <td>Active Params</td>
175
+ <td align="center" style="vertical-align: middle; text-align: center; ">37B</td>
176
+ <td align="center" style="vertical-align: middle; text-align: center; ">23B</td>
177
+ <td align="center" style="vertical-align: middle; text-align: center; ">17B</td>
178
+ <td align="center" style="vertical-align: middle; text-align: center; ">40B</td>
179
+ <td align="center" style="vertical-align: middle; text-align: center; ">49B</td>
180
+ </tr>
181
+ <tr>
182
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>World Knowledge</i></td>
183
+ </tr>
184
+ <tr>
185
+ <td>MMLU-Pro</td>
186
+ <td align="center" style="vertical-align: middle; text-align: center; ">83.5</td>
187
+ <td align="center" style="vertical-align: middle; text-align: center; ">83.8</td>
188
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.8</td>
189
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.0</td>
190
+ <td align="center" style="vertical-align: middle; text-align: center; ">87.5</td>
191
+ </tr>
192
+ <tr>
193
+ <td>GPQA-Diamond</td>
194
+ <td align="center" style="vertical-align: middle; text-align: center; ">82.2</td>
195
+ <td align="center" style="vertical-align: middle; text-align: center; ">79.1</td>
196
+ <td align="center" style="vertical-align: middle; text-align: center; ">88.4</td>
197
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.2</td>
198
+ <td align="center" style="vertical-align: middle; text-align: center; ">90.1</td>
199
+ </tr>
200
+ <tr>
201
+ <td>Humanity's Last Exam</td>
202
+ <td align="center" style="vertical-align: middle; text-align: center; ">18.3</td>
203
+ <td align="center" style="vertical-align: middle; text-align: center; ">13.6</td>
204
+ <td align="center" style="vertical-align: middle; text-align: center; ">28.7</td>
205
+ <td align="center" style="vertical-align: middle; text-align: center; ">31.0</td>
206
+ <td align="center" style="vertical-align: middle; text-align: center; ">37.7</td>
207
+ </tr>
208
+ <tr>
209
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Math</i></td>
210
+ </tr>
211
+ <tr>
212
+ <td>AIME 2026</td>
213
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.3</td>
214
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.2</td>
215
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.3</td>
216
+ <td align="center" style="vertical-align: middle; text-align: center; ">95.3</td>
217
+ <td align="center" style="vertical-align: middle; text-align: center; ">95.2</td>
218
+ </tr>
219
+ <tr>
220
+ <td>HMMT Feb 2026</td>
221
+ <td align="center" style="vertical-align: middle; text-align: center; ">78.4</td>
222
+ <td align="center" style="vertical-align: middle; text-align: center; ">80.7</td>
223
+ <td align="center" style="vertical-align: middle; text-align: center; ">84.6</td>
224
+ <td align="center" style="vertical-align: middle; text-align: center; ">82.6</td>
225
+ <td align="center" style="vertical-align: middle; text-align: center; ">95.2</td>
226
+ </tr>
227
+ <tr>
228
+ <td>IMO Answer</td>
229
+ <td align="center" style="vertical-align: middle; text-align: center; ">78.6</td>
230
+ <td align="center" style="vertical-align: middle; text-align: center; ">76.3</td>
231
+ <td align="center" style="vertical-align: middle; text-align: center; ">80.9</td>
232
+ <td align="center" style="vertical-align: middle; text-align: center; ">83.8</td>
233
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.8</td>
234
+ </tr>
235
+ <tr>
236
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Coding / Agentic Coding</i></td>
237
+ </tr>
238
+ <tr>
239
+ <td>SciCode</td>
240
+ <td align="center" style="vertical-align: middle; text-align: center; ">37.4</td>
241
+ <td align="center" style="vertical-align: middle; text-align: center; ">35.6</td>
242
+ <td align="center" style="vertical-align: middle; text-align: center; ">42.0</td>
243
+ <td align="center" style="vertical-align: middle; text-align: center; ">43.8</td>
244
+ <td align="center" style="vertical-align: middle; text-align: center; ">50.0</td>
245
+ </tr>
246
+ <tr>
247
+ <td>SWE Bench Verified</td>
248
+ <td align="center" style="vertical-align: middle; text-align: center; ">68.2</td>
249
+ <td align="center" style="vertical-align: middle; text-align: center; ">49.4</td>
250
+ <td align="center" style="vertical-align: middle; text-align: center; ">76.4</td>
251
+ <td align="center" style="vertical-align: middle; text-align: center; ">73.6</td>
252
+ <td align="center" style="vertical-align: middle; text-align: center; ">80.6</td>
253
+ </tr>
254
+ <tr>
255
+ <td>Terminal-Bench 2.1</td>
256
+ <td align="center" style="vertical-align: middle; text-align: center; ">43.8</td>
257
+ <td align="center" style="vertical-align: middle; text-align: center; ">30.3</td>
258
+ <td align="center" style="vertical-align: middle; text-align: center; ">51.3</td>
259
+ <td align="center" style="vertical-align: middle; text-align: center; ">61.8</td>
260
+ <td align="center" style="vertical-align: middle; text-align: center; ">64.0</td>
261
+ </tr>
262
+ <tr>
263
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Agentic Tool Use</i></td>
264
+ </tr>
265
+ <tr>
266
+ <td>τ<sup>3</sup>-Banking</td>
267
+ <td align="center" style="vertical-align: middle; text-align: center; ">14.2</td>
268
+ <td align="center" style="vertical-align: middle; text-align: center; ">14.2</td>
269
+ <td align="center" style="vertical-align: middle; text-align: center; ">13.4</td>
270
+ <td align="center" style="vertical-align: middle; text-align: center; ">11.5</td>
271
+ <td align="center" style="vertical-align: middle; text-align: center; ">25.8</td>
272
+ </tr>
273
+ <tr>
274
+ <td>Claw-Eval</td>
275
+ <td align="center" style="vertical-align: middle; text-align: center; ">77.7</td>
276
+ <td align="center" style="vertical-align: middle; text-align: center; ">70.3</td>
277
+ <td align="center" style="vertical-align: middle; text-align: center; ">79.7</td>
278
+ <td align="center" style="vertical-align: middle; text-align: center; ">84.4</td>
279
+ <td align="center" style="vertical-align: middle; text-align: center; ">82.7</td>
280
+ </tr>
281
+ <tr>
282
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Instruction Following</i></td>
283
+ </tr>
284
+ <tr>
285
+ <td>IFEval</td>
286
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.4</td>
287
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.7</td>
288
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.6</td>
289
+ <td align="center" style="vertical-align: middle; text-align: center; ">93.9</td>
290
+ <td align="center" style="vertical-align: middle; text-align: center; ">94.0</td>
291
+ </tr>
292
+ <tr>
293
+ <td>IFBench</td>
294
+ <td align="center" style="vertical-align: middle; text-align: center; ">72.6</td>
295
+ <td align="center" style="vertical-align: middle; text-align: center; ">67.3</td>
296
+ <td align="center" style="vertical-align: middle; text-align: center; ">76.5</td>
297
+ <td align="center" style="vertical-align: middle; text-align: center; ">76.3</td>
298
+ <td align="center" style="vertical-align: middle; text-align: center; ">76.5</td>
299
+ </tr>
300
+ <tr>
301
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Long Context Understanding</i></td>
302
+ </tr>
303
+ <tr>
304
+ <td>OpenAI-MRCR</td>
305
+ <td align="center" style="vertical-align: middle; text-align: center; ">94.4</td>
306
+ <td align="center" style="vertical-align: middle; text-align: center; ">52.3</td>
307
+ <td align="center" style="vertical-align: middle; text-align: center; ">93.0</td>
308
+ <td align="center" style="vertical-align: middle; text-align: center; ">71.5</td>
309
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.9</td>
310
+ </tr>
311
+ <tr>
312
+ <td>AA-LCR</td>
313
+ <td align="center" style="vertical-align: middle; text-align: center; ">56.2</td>
314
+ <td align="center" style="vertical-align: middle; text-align: center; ">53.5</td>
315
+ <td align="center" style="vertical-align: middle; text-align: center; ">65.7</td>
316
+ <td align="center" style="vertical-align: middle; text-align: center; ">62.3</td>
317
+ <td align="center" style="vertical-align: middle; text-align: center; ">66.3</td>
318
+ </tr>
319
+ <tr>
320
+ <td>Ko-LongBench</td>
321
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.6</td>
322
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.8</td>
323
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.3</td>
324
+ <td align="center" style="vertical-align: middle; text-align: center; ">83.6</td>
325
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.4</td>
326
+ </tr>
327
+ <tr>
328
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Korean</i></td>
329
+ </tr>
330
+ <tr>
331
+ <td>KMMLU-Pro</td>
332
+ <td align="center" style="vertical-align: middle; text-align: center; ">69.1</td>
333
+ <td align="center" style="vertical-align: middle; text-align: center; ">67.3</td>
334
+ <td align="center" style="vertical-align: middle; text-align: center; ">77.4</td>
335
+ <td align="center" style="vertical-align: middle; text-align: center; ">75.8</td>
336
+ <td align="center" style="vertical-align: middle; text-align: center; ">80.5</td>
337
+ </tr>
338
+ <tr>
339
+ <td>Click</td>
340
+ <td align="center" style="vertical-align: middle; text-align: center; ">84.2</td>
341
+ <td align="center" style="vertical-align: middle; text-align: center; ">83.9</td>
342
+ <td align="center" style="vertical-align: middle; text-align: center; ">88.9</td>
343
+ <td align="center" style="vertical-align: middle; text-align: center; ">88.7</td>
344
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.6</td>
345
+ </tr>
346
+ <tr>
347
+ <td>HRM8K-KSM</td>
348
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.1</td>
349
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.9</td>
350
+ <td align="center" style="vertical-align: middle; text-align: center; ">91.2</td>
351
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.4</td>
352
+ <td align="center" style="vertical-align: middle; text-align: center; ">94.3</td>
353
+ </tr>
354
+ <tr>
355
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Multilinguality</i></td>
356
+ </tr>
357
+ <tr>
358
+ <td>MMMLU</td>
359
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.6</td>
360
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.2</td>
361
+ <td align="center" style="vertical-align: middle; text-align: center; ">90.6</td>
362
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.7</td>
363
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.6</td>
364
+ </tr>
365
+ <tr>
366
+ <td>GlobalMMLU-Lite</td>
367
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.6</td>
368
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.9</td>
369
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.1</td>
370
+ <td align="center" style="vertical-align: middle; text-align: center; ">90.7</td>
371
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.0</td>
372
+ </tr>
373
+ <tr>
374
+ <td>PolyMath</td>
375
+ <td align="center" style="vertical-align: middle; text-align: center; ">71.3</td>
376
+ <td align="center" style="vertical-align: middle; text-align: center; ">57.4</td>
377
+ <td align="center" style="vertical-align: middle; text-align: center; ">73.3</td>
378
+ <td align="center" style="vertical-align: middle; text-align: center; ">73.8</td>
379
+ <td align="center" style="vertical-align: middle; text-align: center; ">80.9</td>
380
+ </tr>
381
+ <tr>
382
+ <td align="center" style="vertical-align: middle; text-align: center; background: linear-gradient(90deg, rgba(252,146,108,0.3) 0%, rgba(227,67,189,0.3) 50%, rgba(104,76,244,0.3) 100%); font-weight: bold; height:32px; padding-top:2px; padding-bottom:2px;" colspan='7'><i>Safety</i></td>
383
+ </tr>
384
+ <tr>
385
+ <td>KGC-Safety</td>
386
+ <td align="center" style="vertical-align: middle; text-align: center; ">99.8</td>
387
+ <td align="center" style="vertical-align: middle; text-align: center; ">96.1</td>
388
+ <td align="center" style="vertical-align: middle; text-align: center; ">92.0</td>
389
+ <td align="center" style="vertical-align: middle; text-align: center; ">69.3</td>
390
+ <td align="center" style="vertical-align: middle; text-align: center; ">82.8</td>
391
+ </tr>
392
+ <tr>
393
+ <td>ROK-Fortress</td>
394
+ <td align="center" style="vertical-align: middle; text-align: center; ">89.5</td>
395
+ <td align="center" style="vertical-align: middle; text-align: center; ">60.9</td>
396
+ <td align="center" style="vertical-align: middle; text-align: center; ">86.1</td>
397
+ <td align="center" style="vertical-align: middle; text-align: center; ">73.2</td>
398
+ <td align="center" style="vertical-align: middle; text-align: center; ">47.6</td>
399
+ </tr>
400
+ </table>
401
+
402
+
403
+
404
+ ## Quickstart
405
+
406
+ ### Serving K-EXAONE 2.0
407
+
408
+ ### SGLang
409
+
410
+ Install SGLang from [our fork](https://github.com/lkm2835/sglang/tree/add-k-exaone2) as follows:
411
+ ```bash
412
+ uv venv
413
+ source .venv/bin/activate
414
+ uv pip install git+https://github.com/lkm2835/sglang@add-k-exaone2
415
+ uv pip install git+https://github.com/nuxlear/transformers@add-k-exaone2
416
+ ```
417
+
418
+ The following script shows how to serve K-EXAONE 2.0 with SGLang on two nodes of 8 x NVIDIA H200 GPUs.
419
+ Before starting the server, set `$HEAD_ADDR` to the IP address of the head (rank-0) node including the port number, and set `$NODE_RANK` to the number index of the node.
420
+
421
+ Run the one of the following scripts on every node, according to your purpose.
422
+
423
+ - Low latency
424
+ ```bash
425
+ sglang serve \
426
+ --model-path LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark \
427
+ --served-model-name K-EXAONE-2.0-750B-A37B-DSpark \
428
+ --tp 16 \
429
+ --dist-init-addr $HEAD_ADDR \
430
+ --nnodes 2 \
431
+ --node-rank $NODE_RANK \
432
+ --reasoning-parser qwen3 \
433
+ --tool-call-parser qwen3_coder \
434
+ --host 0.0.0.0 \
435
+ --port 8000 \
436
+ --max-running-requests 128 \
437
+ --speculative-algorithm DSPARK \
438
+ --speculative-num-draft-tokens 8 \
439
+ --mem-fraction-static 0.85 \
440
+ --swa-full-tokens-ratio 0.6
441
+ ```
442
+
443
+ - High throughput
444
+ ```bash
445
+ sglang serve \
446
+ --model-path LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark \
447
+ --served-model-name K-EXAONE-2.0-750B-A37B-DSpark \
448
+ --tp 16 \
449
+ --dp 2 \
450
+ --enable-dp-attention \
451
+ --enable-dp-lm-head \
452
+ --dist-init-addr $HEAD_ADDR \
453
+ --nnodes 2 \
454
+ --node-rank $NODE_RANK \
455
+ --reasoning-parser qwen3 \
456
+ --tool-call-parser qwen3_coder \
457
+ --host 0.0.0.0 \
458
+ --port 8000 \
459
+ --max-running-requests 192 \
460
+ --cuda-graph-max-bs 96 \
461
+ --speculative-algorithm DSPARK \
462
+ --speculative-num-draft-tokens 8 \
463
+ --mem-fraction-static 0.85 \
464
+ --swa-full-tokens-ratio 0.6
465
+ ```
466
+
467
+
468
+ > [!NOTE]
469
+ > If you are using NVIDIA B200 GPUs, please set `EXAONE_ENABLE_SWA_KV=0` and add the `--disable-prefill-cuda-graph` option to prevent issues with model generation collapse.
470
+ > We will update this note once these issues have been resolved.
471
+
472
+ ### vLLM
473
+
474
+ You should install the vLLM library from [our fork](https://github.com/lkm2835/vllm/tree/add-k-exaone2) as below:
475
+ ```bash
476
+ uv venv
477
+ source .venv/bin/activate
478
+ uv pip install git+https://github.com/lkm2835/vllm@add-k-exaone2 --torch-backend auto
479
+ uv pip install git+https://github.com/nuxlear/transformers@add-k-exaone2
480
+ ```
481
+
482
+ The following script shows how to serve K-EXAONE 2.0 with vLLM on two nodes of 8 x NVIDIA H200 GPUs.
483
+ Before starting the server, set `$HEAD_IP` to the IP address of the head (rank-0) node, and set `$NODE_RANK` to the number index of the node.
484
+
485
+ Run the below script on every server.
486
+ ```bash
487
+ HEADLESS_ARG=''
488
+ if [ \"\$NODE_RANK\" -ne 0 ]; then
489
+ HEADLESS_ARG='--headless'
490
+ fi
491
+
492
+ exec vllm serve LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark \
493
+ --served-model-name K-EXAONE-2.0-750B-A37B-DSpark \
494
+ --trust-remote-code \
495
+ --tensor-parallel-size 16 \
496
+ --distributed-executor-backend mp \
497
+ --nnodes 2 \
498
+ --node-rank $NODE_RANK \
499
+ --master-addr $HEAD_IP \
500
+ --master-addr 30000 \
501
+ --gpu-memory-utilization 0.9 \
502
+ --max-num-seqs 256 \
503
+ --reasoning-parser qwen3 \
504
+ --enable-auto-tool-choice \
505
+ --tool-call-parser qwen3_xml \
506
+ --host 0.0.0.0 \
507
+ --port 8000 \
508
+ --speculative_config '{
509
+ "method": "mtp",
510
+ "num_speculative_tokens": 4
511
+ }' \
512
+ $HEADLESS_ARG
513
+ ```
514
+
515
+ > [!NOTE]
516
+ > Currently, serving K-EXAONE 2.0 with DSpark is not supported on vLLM.
517
+ > We will update this note once these issues have been resolved.
518
+
519
+
520
+
521
+ ### Using K-EXAONE 2.0
522
+
523
+ > [!IMPORTANT]
524
+ > To achieve the expected performance, we recommend using the following configurations:
525
+ > - We recommend using `temperature=1.0` and `top_p=0.95` for better output quality in most cases.
526
+ > - K-EXAONE 2.0 uses `enable_thinking=True` by default. Thus, you need to set `enable_thinking=False` to use non-reasoning mode.
527
+ > - We recommend using `preserve_thinking=True` in long-running tasks, such as agentic use or deep research.
528
+
529
+
530
+ Once K-EXAONE 2.0 is running on an inference engine, you can access it using the OpenAI Python SDK.
531
+
532
+ #### Reasoning mode
533
+
534
+ For tasks that require high accuracy, you can use the K-EXAONE 2.0 model in reasoning mode. The K-EXAONE 2.0 model supports the `preserve_thinking` option, which allows it to track previous `reasoning_content` across subsequent conversations. For agentic workflows, it is recommended to use `preserve_thinking=True`.
535
+
536
+ ```python
537
+ from openai import OpenAI
538
+
539
+ client = OpenAI(
540
+ base_url="http://localhost:8000/v1",
541
+ api_key="EMPTY",
542
+ )
543
+
544
+ messages = [
545
+ {
546
+ "role": "user",
547
+ "content": "Implement fibonacci with python code.",
548
+ }
549
+ ]
550
+
551
+ response = client.chat.completions.create(
552
+ model="LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark",
553
+ messages=messages,
554
+ max_tokens=32768,
555
+ temperature=1.0,
556
+ top_p=0.95,
557
+ extra_body={
558
+ "chat_template_kwargs": {
559
+ "enable_thinking": True, # default: True
560
+ "preserve_thinking": True, # default: False
561
+ }
562
+ },
563
+ )
564
+ print(response)
565
+ ```
566
+ #### Non-reasoning mode
567
+
568
+ For tasks where latency matters more than accuracy, you can run the K-EXAONE 2.0 model in non-reasoning mode.
569
+
570
+ ```python
571
+ from openai import OpenAI
572
+
573
+ client = OpenAI(
574
+ base_url="http://localhost:8000/v1",
575
+ api_key="EMPTY",
576
+ )
577
+
578
+ messages = [
579
+ {
580
+ "role": "user",
581
+ "content": "Explain how useful you are.",
582
+ }
583
+ ]
584
+
585
+ response = client.chat.completions.create(
586
+ model="LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark",
587
+ messages=messages,
588
+ max_tokens=32768,
589
+ temperature=1.0,
590
+ top_p=0.95,
591
+ extra_body={
592
+ "chat_template_kwargs": {
593
+ "enable_thinking": False, # default: True
594
+ "preserve_thinking": False, # default: False
595
+ }
596
+ },
597
+ )
598
+ print(response)
599
+ ```
600
+ #### Tool calling
601
+
602
+ For your AI-powered agent, you can leverage K-EXAONE 2.0’s tool calling capability.
603
+
604
+ ```python
605
+ from openai import OpenAI
606
+
607
+ client = OpenAI(
608
+ base_url="http://localhost:8000/v1",
609
+ api_key="EMPTY",
610
+ )
611
+
612
+ tools = [
613
+ {
614
+ "type": "function",
615
+ "function": {
616
+ "name": "roll_dice",
617
+ "description": "Roll the dice with the number 1 to N. User can select the number N.",
618
+ "parameters": {
619
+ "type": "object",
620
+ "properties": {
621
+ "max_num": {
622
+ "type": "integer",
623
+ "description": "The maximum number on the dice."
624
+ }
625
+ },
626
+ "required": ["max_num"]
627
+ },
628
+ },
629
+ }
630
+ ]
631
+
632
+ messages = [
633
+ {
634
+ "role": "user",
635
+ "content": "Roll a D20 twice and sum the results."
636
+ }
637
+ ]
638
+
639
+ response = client.chat.completions.create(
640
+ model="LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark",
641
+ messages=messages,
642
+ tools=tools,
643
+ max_tokens=32768,
644
+ temperature=1.0,
645
+ top_p=0.95,
646
+ extra_body={
647
+ "chat_template_kwargs": {
648
+ "enable_thinking": True, # default: True
649
+ "preserve_thinking": True, # default: False
650
+ }
651
+ },
652
+ )
653
+
654
+ print(response)
655
+ ```
656
+
657
+ ### Agentic Use
658
+
659
+ You can leverage K-EXAONE 2.0’s agentic capabilities by integrating it with agent frameworks and harnesses.
660
+
661
+ #### OpenCode
662
+
663
+ To use the deployed K-EXAONE 2.0 model, you may need to update your `opencode.json` file. Below is an example JSON configuration for setting up a custom server as your model provider.
664
+
665
+ ```json
666
+ {
667
+ "$schema": "https://opencode.ai/config.json",
668
+ "provider": {
669
+ "local": {
670
+ "npm": "@ai-sdk/openai-compatible",
671
+ "name": "Local OpenAI-compatible server",
672
+ "options": {
673
+ "baseURL": "http://localhost:8000/v1",
674
+ "extraBody": {
675
+ "chat_template_kwargs": {
676
+ "enable_thinking": true,
677
+ "preserve_thinking": true
678
+ }
679
+ }
680
+ },
681
+ "models": {
682
+ "K-EXAONE-2": {
683
+ "name": "K-EXAONE 2.0",
684
+ "limit": {
685
+ "context": 262144,
686
+ "output": 32768
687
+ }
688
+ }
689
+ }
690
+ }
691
+ }
692
+ }
693
+ ```
694
+
695
+
696
+
697
+
698
+ ## Limitation
699
+
700
+ K-EXAONE 2.0 language models, like all existing language models, have certain limitations and may occasionally generate inappropriate responses. The language model generates responses based on the output probability of tokens, and it is determined during learning from training data. While we make every effort to exclude personal, harmful, and biased information from the training data, some problematic content may still be included, potentially leading to undesirable responses. Please note that the text generated by K-EXAONE 2.0 language models does not reflect the views of LG AI Research.
701
+ - Inappropriate answers may be generated, which contain personal, harmful or other inappropriate information.
702
+ - Biased responses may be generated, which are associated with age, gender, race, and so on.
703
+ - The generated responses rely heavily on statistics from the training data, which can result in the generation of semantically or syntactically incorrect sentences.
704
+ - Since the models do not reflect the latest information, the responses may be false or contradictory.
705
+
706
+ LG AI Research strives to reduce potential risks that may arise from K-EXAONE 2.0 language models. Users are not allowed to engage in any malicious activities (e.g., keying in illegal information) that may induce the creation of inappropriate outputs violating LG AI’s ethical principles when using K-EXAONE 2.0 language models.
707
+
708
+ ## License
709
+
710
+ The model is licensed under [Apache License 2.0](https://huggingface.co/LGAI-EXAONE/K-EXAONE-2.0-750B-A37B-DSpark/blob/main/LICENSE).
711
+
712
+
713
+ ## Citation
714
+
715
+ ```
716
+ @article{k-exaone-2.0,
717
+ title={K-EXAONE 2.0 Technical Report},
718
+ author={{LG AI Research}},
719
+ journal={arXiv preprint arXiv:XXXX.XXXXX},
720
+ year={2026}
721
+ }
722
+ ```
723
+
724
+
725
+ ## Contact
726
+
727
+ LG AI Research Technical Support: contact_us@lgresearch.ai
728
+
assets/K-EXAONE-2.0-Technical-Report.pdf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78c1bc2a9ccf3de222e5cada78922233080f4dc0fbc9f3998e0d4018aaa085cf
3
+ size 2952264
assets/K-EXAONE_logo_gray.png ADDED

Git LFS Details

  • SHA256: c51e9ddeac2cf198d690c3307db53b265801c1f0ca190ce0d4e558d9f27a46d7
  • Pointer size: 131 Bytes
  • Size of remote file: 499 kB
assets/k-ai-logo-no-bg.png ADDED

Git LFS Details

  • SHA256: 50c193a9465c2c724d2feb70cb6b89d37b5d0fec53c93ec6db23a1e9a10419b1
  • Pointer size: 132 Bytes
  • Size of remote file: 3.22 MB
assets/main_figure.png ADDED

Git LFS Details

  • SHA256: ddfacc7dbff499377ff833bcef3dc1a794089c7051c7911e13703b9d6043d941
  • Pointer size: 132 Bytes
  • Size of remote file: 1.87 MB
chat_template.jinja ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {% set image_count = namespace(value=0) %}
2
+ {% set video_count = namespace(value=0) %}
3
+
4
+ {%- set role_indicators = {
5
+ 'user': '<|user|>\n',
6
+ 'assistant': '<|assistant|>\n',
7
+ 'system': '<|system|>\n',
8
+ 'tool': '<|tool|>\n',
9
+ 'tool_declare': '<|tool_declare|>\n'
10
+ } %}
11
+ {%- set end_of_turn = '<|endofturn|>\n' %}
12
+
13
+
14
+ {%- macro declare_available_tools(tools) %}
15
+ {{- "# Tools\n" }}
16
+ {{- "The available tools are defined below in JSON format.\n" }}
17
+ {{- "When calling a tool, use XML with <function=...> and one <parameter=...> block per argument.\n" }}
18
+ {{- "\n" }}
19
+ {%- for tool in tools %}
20
+ {{- "<tool>" }}
21
+ {{- tool | tojson(ensure_ascii=False) | safe }}
22
+ {{- "</tool>\n" }}
23
+ {%- endfor %}
24
+ {{- "\n# Tool Call Format\n" }}
25
+ {{- "<tool_call>\n" }}
26
+ {{- "<function=example_tool_name>\n" }}
27
+ {{- "<parameter=arg1>\n" }}
28
+ {{- "value1\n" }}
29
+ {{- "</parameter>\n" }}
30
+ {{- "<parameter=arg2>\n" }}
31
+ {{- "2\n" }}
32
+ {{- "</parameter>\n" }}
33
+ {{- "</function>\n" }}
34
+ {{- "</tool_call>" }}
35
+ {%- endmacro %}
36
+
37
+
38
+ {%- macro render_tool_call(tool_call, arguments) %}
39
+ {{- "<tool_call>\n" }}
40
+ {{- "<function=" }}{{- tool_call.name }}{{- ">\n" }}
41
+ {%- for args_name, args_value in arguments.items() %}
42
+ {{- "<parameter=" }}{{- args_name }}{{- ">\n" }}
43
+ {%- if args_value is string %}
44
+ {{- args_value }}
45
+ {%- else %}
46
+ {{- args_value | tojson(ensure_ascii=False) | safe }}
47
+ {%- endif %}
48
+ {{- "\n</parameter>\n" }}
49
+ {%- endfor %}
50
+ {{- "</function>\n" }}
51
+ {{- "</tool_call>" }}
52
+ {%- endmacro %}
53
+
54
+
55
+ {%- macro render_tool_response(msg) %}
56
+ {{- "<tool_result>" }}
57
+ {%- if msg.content is defined %}
58
+ {%- if msg.content is string %}
59
+ {{- msg.content }}
60
+ {%- else %}
61
+ {{- msg.content | tojson(ensure_ascii=False) | safe }}
62
+ {%- endif %}
63
+ {%- endif %}
64
+ {{- "</tool_result>" }}
65
+ {%- endmacro %}
66
+
67
+
68
+ {%- set ns = namespace(last_query_index = messages|length - 1, last_query_index_not_yet_determined = true) %}
69
+ {%- for message in messages[::-1] %}
70
+ {%- set index = (messages|length - 1) - loop.index0 %}
71
+ {%- if ns.last_query_index_not_yet_determined and message.role == "user" and message.content is string %}
72
+ {%- set ns.last_query_index = index -%}
73
+ {%- set ns.last_query_index_not_yet_determined = false -%}
74
+ {%- endif %}
75
+ {%- endfor %}
76
+
77
+ {%- if tools is defined and tools %}
78
+ {{- role_indicators['tool_declare'] }}
79
+ {{- declare_available_tools(tools) }}
80
+ {{- end_of_turn -}}
81
+ {%- endif %}
82
+
83
+ {%- for i in range(messages | length) %}
84
+ {%- set msg = messages[i] %}
85
+ {%- set role = msg.role %}
86
+ {%- if role not in role_indicators %}
87
+ {{- raise_exception('Unknown role: ' ~ role) }}
88
+ {%- endif %}
89
+
90
+ {%- if i == 0 %}
91
+ {%- if role == 'system' %}
92
+ {{- role_indicators['system'] }}
93
+ {{- msg.content }}
94
+ {{- end_of_turn -}}
95
+ {%- continue %}
96
+ {%- endif %}
97
+ {%- endif %}
98
+
99
+ {%- if role == 'assistant' %}
100
+ {{- role_indicators['assistant'] }}
101
+
102
+ {%- set content = (msg.content if (msg.content is defined and msg.content) else "") -%}
103
+ {%- set reasoning = none -%}
104
+
105
+ {%- if msg.reasoning_content is defined and msg.reasoning_content %}
106
+ {%- set reasoning = msg.reasoning_content.strip() -%}
107
+ {%- elif content and "</think>" in content %}
108
+ {%- set _parts = content.split('</think>') -%}
109
+ {%- set reasoning = _parts[0].lstrip('<think>').strip() -%}
110
+ {%- set content = _parts[-1].strip() -%}
111
+ {%- endif %}
112
+
113
+ {%- if not (reasoning and ((preserve_thinking is defined and preserve_thinking) or i > ns.last_query_index))%}
114
+ {%- set reasoning = none %}
115
+ {%- endif %}
116
+
117
+ {%- if content is string %}
118
+ {%- set content = content.strip() -%}
119
+ {%- endif %}
120
+
121
+ {{- "<think>\n" }}
122
+ {{- (reasoning if reasoning is not none else "") }}
123
+ {{- "\n</think>\n\n" }}
124
+
125
+ {{- content }}
126
+
127
+ {%- if msg.tool_calls %}
128
+ {%- if content is defined and content %}
129
+ {{- "\n" }}
130
+ {%- endif %}
131
+ {%- for tool_call in msg.tool_calls %}
132
+ {%- if tool_call.function is defined %}
133
+ {%- set tool_call = tool_call.function %}
134
+ {%- endif %}
135
+
136
+ {%- if tool_call.arguments is defined %}
137
+ {%- set arguments = tool_call.arguments %}
138
+ {%- elif tool_call.parameters is defined %}
139
+ {%- set arguments = tool_call.parameters %}
140
+ {%- else %}
141
+ {{- raise_exception('arguments or parameters are mandatory: ' ~ tool_call) }}
142
+ {%- endif %}
143
+
144
+ {{- render_tool_call(tool_call, arguments) }}
145
+ {%- if not loop.last %}
146
+ {{- "\n" }}
147
+ {%- endif %}
148
+ {%- endfor %}
149
+ {%- endif %}
150
+ {{- end_of_turn -}}
151
+
152
+ {%- elif role == "tool" %}
153
+ {%- if i == 0 or messages[i - 1].role != "tool" %}
154
+ {{- role_indicators['tool'] }}
155
+ {%- endif %}
156
+ {{- render_tool_response(msg) }}
157
+ {%- if loop.last or messages[i + 1].role != "tool" %}
158
+ {{- end_of_turn -}}
159
+ {%- else %}
160
+ {{- "\n" }}
161
+ {%- endif %}
162
+
163
+ {%- else %}
164
+ {{- role_indicators[role] }}
165
+ {%- if msg.content is string %}
166
+ {{- msg.content }}
167
+ {%- else %}
168
+ {%- for content in msg.content %}
169
+ {%- if content.type == 'image' %}
170
+ {%- set image_count.value = image_count.value + 1 %}
171
+ {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif %}<vision><image_pad></vision>
172
+ {%- elif content.type == 'video' %}
173
+ {%- set video_count.value = video_count.value + 1 %}
174
+ {%- if add_vision_id %}Video {{ video_count.value }}: {% endif %}<vision><video_pad></vision>
175
+ {%- elif content.type == 'text' %}
176
+ {{- content.text }}
177
+ {%- else %}
178
+ {{- content.text }}
179
+ {%- endif %}
180
+ {%- endfor %}
181
+ {%- endif %}
182
+ {{- end_of_turn -}}
183
+ {%- endif %}
184
+ {% endfor %}
185
+
186
+
187
+ {%- if add_generation_prompt %}
188
+ {{- role_indicators['assistant'] }}
189
+ {%- if enable_thinking is not defined or enable_thinking is true %}
190
+ {{- "<think>\n" }}
191
+ {%- else %}
192
+ {{- "<think>\n\n</think>\n\n" }}
193
+ {%- endif %}
194
+ {%- endif %}
config.json ADDED
@@ -0,0 +1,391 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "ExaoneMoeForCausalLM"
4
+ ],
5
+ "attention_dropout": 0.0,
6
+ "block_size": 7,
7
+ "bos_token_id": 1,
8
+ "confidence_head_with_markov": true,
9
+ "dtype": "bfloat16",
10
+ "enable_confidence_head": true,
11
+ "eos_token_id": 53,
12
+ "head_dim": 128,
13
+ "hidden_act": "silu",
14
+ "hidden_size": 6144,
15
+ "initializer_range": 0.02,
16
+ "intermediate_size": 18432,
17
+ "layer_types": [
18
+ "full_attention",
19
+ "sliding_attention",
20
+ "sliding_attention",
21
+ "sliding_attention",
22
+ "sliding_attention",
23
+ "full_attention",
24
+ "sliding_attention",
25
+ "sliding_attention",
26
+ "sliding_attention",
27
+ "full_attention",
28
+ "sliding_attention",
29
+ "sliding_attention",
30
+ "sliding_attention",
31
+ "full_attention",
32
+ "sliding_attention",
33
+ "sliding_attention",
34
+ "sliding_attention",
35
+ "full_attention",
36
+ "sliding_attention",
37
+ "sliding_attention",
38
+ "sliding_attention",
39
+ "full_attention",
40
+ "sliding_attention",
41
+ "sliding_attention",
42
+ "sliding_attention",
43
+ "full_attention",
44
+ "sliding_attention",
45
+ "sliding_attention",
46
+ "sliding_attention",
47
+ "full_attention",
48
+ "sliding_attention",
49
+ "sliding_attention",
50
+ "sliding_attention",
51
+ "full_attention",
52
+ "sliding_attention",
53
+ "sliding_attention",
54
+ "sliding_attention",
55
+ "full_attention",
56
+ "sliding_attention",
57
+ "sliding_attention",
58
+ "sliding_attention",
59
+ "full_attention",
60
+ "sliding_attention",
61
+ "sliding_attention",
62
+ "sliding_attention",
63
+ "full_attention",
64
+ "sliding_attention",
65
+ "sliding_attention",
66
+ "sliding_attention",
67
+ "full_attention",
68
+ "sliding_attention",
69
+ "sliding_attention",
70
+ "sliding_attention",
71
+ "full_attention",
72
+ "sliding_attention",
73
+ "sliding_attention",
74
+ "sliding_attention",
75
+ "full_attention",
76
+ "sliding_attention",
77
+ "sliding_attention",
78
+ "sliding_attention",
79
+ "full_attention",
80
+ "sliding_attention",
81
+ "sliding_attention",
82
+ "sliding_attention",
83
+ "full_attention",
84
+ "sliding_attention",
85
+ "sliding_attention",
86
+ "sliding_attention",
87
+ "full_attention",
88
+ "sliding_attention",
89
+ "sliding_attention",
90
+ "sliding_attention",
91
+ "full_attention",
92
+ "sliding_attention",
93
+ "sliding_attention",
94
+ "sliding_attention",
95
+ "full_attention"
96
+ ],
97
+ "markov_head_type": "vanilla",
98
+ "markov_rank": 256,
99
+ "max_position_embeddings": 262144,
100
+ "mlp_layer_types": [
101
+ "dense",
102
+ "dense",
103
+ "sparse",
104
+ "sparse",
105
+ "sparse",
106
+ "sparse",
107
+ "sparse",
108
+ "sparse",
109
+ "sparse",
110
+ "sparse",
111
+ "sparse",
112
+ "sparse",
113
+ "sparse",
114
+ "sparse",
115
+ "sparse",
116
+ "sparse",
117
+ "sparse",
118
+ "sparse",
119
+ "sparse",
120
+ "sparse",
121
+ "sparse",
122
+ "sparse",
123
+ "sparse",
124
+ "sparse",
125
+ "sparse",
126
+ "sparse",
127
+ "sparse",
128
+ "sparse",
129
+ "sparse",
130
+ "sparse",
131
+ "sparse",
132
+ "sparse",
133
+ "sparse",
134
+ "sparse",
135
+ "sparse",
136
+ "sparse",
137
+ "sparse",
138
+ "sparse",
139
+ "sparse",
140
+ "sparse",
141
+ "sparse",
142
+ "sparse",
143
+ "sparse",
144
+ "sparse",
145
+ "sparse",
146
+ "sparse",
147
+ "sparse",
148
+ "sparse",
149
+ "sparse",
150
+ "sparse",
151
+ "sparse",
152
+ "sparse",
153
+ "sparse",
154
+ "sparse",
155
+ "sparse",
156
+ "sparse",
157
+ "sparse",
158
+ "sparse",
159
+ "sparse",
160
+ "sparse",
161
+ "sparse",
162
+ "sparse",
163
+ "sparse",
164
+ "sparse",
165
+ "sparse",
166
+ "sparse",
167
+ "sparse",
168
+ "sparse",
169
+ "sparse",
170
+ "sparse",
171
+ "sparse",
172
+ "sparse",
173
+ "sparse",
174
+ "sparse",
175
+ "sparse",
176
+ "sparse",
177
+ "sparse",
178
+ "sparse"
179
+ ],
180
+ "model_type": "exaone_moe",
181
+ "moe_intermediate_size": 2048,
182
+ "mtp_layer_types": [
183
+ "full_attention",
184
+ "full_attention",
185
+ "full_attention",
186
+ "full_attention",
187
+ "full_attention"
188
+ ],
189
+ "mtp_loss_scaling_factor": 0.05,
190
+ "mtp_num_speculative_steps": 4,
191
+ "mtp_sliding_windows": [
192
+ 0,
193
+ 0,
194
+ 0,
195
+ 0,
196
+ 0
197
+ ],
198
+ "n_group": 1,
199
+ "noise_token_id": 84,
200
+ "norm_topk_prob": true,
201
+ "num_attention_heads": 64,
202
+ "num_experts": 256,
203
+ "num_experts_per_tok": 8,
204
+ "num_hidden_layers": 78,
205
+ "num_key_value_heads": 8,
206
+ "num_nextn_predict_layers": 5,
207
+ "num_shared_experts": 1,
208
+ "num_target_layers": 78,
209
+ "pad_token_id": 0,
210
+ "rms_norm_eps": 1e-05,
211
+ "rope_parameters": {
212
+ "rope_theta": 1000000,
213
+ "rope_type": "default"
214
+ },
215
+ "routed_scaling_factor": 2.5,
216
+ "scoring_func": "sigmoid",
217
+ "sliding_windows": [
218
+ 0,
219
+ 4096,
220
+ 128,
221
+ 128,
222
+ 128,
223
+ 0,
224
+ 128,
225
+ 128,
226
+ 128,
227
+ 0,
228
+ 128,
229
+ 128,
230
+ 128,
231
+ 0,
232
+ 128,
233
+ 128,
234
+ 128,
235
+ 0,
236
+ 128,
237
+ 128,
238
+ 128,
239
+ 0,
240
+ 128,
241
+ 128,
242
+ 128,
243
+ 0,
244
+ 128,
245
+ 128,
246
+ 128,
247
+ 0,
248
+ 128,
249
+ 128,
250
+ 128,
251
+ 0,
252
+ 128,
253
+ 128,
254
+ 128,
255
+ 0,
256
+ 128,
257
+ 128,
258
+ 128,
259
+ 0,
260
+ 128,
261
+ 128,
262
+ 128,
263
+ 0,
264
+ 128,
265
+ 128,
266
+ 128,
267
+ 0,
268
+ 128,
269
+ 128,
270
+ 128,
271
+ 0,
272
+ 128,
273
+ 128,
274
+ 128,
275
+ 0,
276
+ 128,
277
+ 128,
278
+ 128,
279
+ 0,
280
+ 128,
281
+ 128,
282
+ 128,
283
+ 0,
284
+ 128,
285
+ 128,
286
+ 128,
287
+ 0,
288
+ 128,
289
+ 128,
290
+ 128,
291
+ 0,
292
+ 128,
293
+ 128,
294
+ 128,
295
+ 0
296
+ ],
297
+ "swiglu_limits": [
298
+ 0.0,
299
+ 0.0,
300
+ 0.0,
301
+ 0.0,
302
+ 0.0,
303
+ 0.0,
304
+ 0.0,
305
+ 0.0,
306
+ 0.0,
307
+ 0.0,
308
+ 0.0,
309
+ 0.0,
310
+ 0.0,
311
+ 0.0,
312
+ 0.0,
313
+ 0.0,
314
+ 0.0,
315
+ 0.0,
316
+ 0.0,
317
+ 0.0,
318
+ 0.0,
319
+ 0.0,
320
+ 0.0,
321
+ 0.0,
322
+ 0.0,
323
+ 0.0,
324
+ 0.0,
325
+ 0.0,
326
+ 0.0,
327
+ 0.0,
328
+ 0.0,
329
+ 0.0,
330
+ 0.0,
331
+ 0.0,
332
+ 0.0,
333
+ 0.0,
334
+ 0.0,
335
+ 0.0,
336
+ 0.0,
337
+ 0.0,
338
+ 0.0,
339
+ 0.0,
340
+ 0.0,
341
+ 0.0,
342
+ 0.0,
343
+ 0.0,
344
+ 0.0,
345
+ 0.0,
346
+ 0.0,
347
+ 0.0,
348
+ 0.0,
349
+ 0.0,
350
+ 0.0,
351
+ 0.0,
352
+ 0.0,
353
+ 0.0,
354
+ 0.0,
355
+ 0.0,
356
+ 0.0,
357
+ 0.0,
358
+ 0.0,
359
+ 0.0,
360
+ 7.0,
361
+ 7.0,
362
+ 7.0,
363
+ 7.0,
364
+ 7.0,
365
+ 7.0,
366
+ 7.0,
367
+ 7.0,
368
+ 7.0,
369
+ 7.0,
370
+ 7.0,
371
+ 7.0,
372
+ 7.0,
373
+ 7.0,
374
+ 7.0,
375
+ 7.0
376
+ ],
377
+ "target_layer_ids": [
378
+ 2,
379
+ 20,
380
+ 38,
381
+ 56,
382
+ 74
383
+ ],
384
+ "tie_word_embeddings": false,
385
+ "tokenizer_class": "GPT2Tokenizer",
386
+ "topk_group": 1,
387
+ "topk_method": "noaux_tc",
388
+ "transformers_version": "5.9.0",
389
+ "use_cache": true,
390
+ "vocab_size": 153600
391
+ }
generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "eos_token_id": 2,
5
+ "pad_token_id": 0,
6
+ "temperature": 1.0,
7
+ "top_p": 0.95,
8
+ "transformers_version": "5.9.0"
9
+ }
model-00001-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bac5808b3b079a018a25031e54f37f91fc76672e5bfbff42d93c8c5a9fbfdf7b
3
+ size 94841171064
model-00002-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:08c4a772592636005163c3d2b18b581f2d4a2d3d6a30c8ff4cc312a41fcabd7e
3
+ size 98465059624
model-00003-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ae9898c329d5f476f0373524675123c880620299be600397a51219bf58d2e70
3
+ size 98465062248
model-00004-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9baa69c7dcbc590aa6771bcf43b5fe72ed5f34ad27d9a50f8caa25fce3d8871d
3
+ size 98465062248
model-00005-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:68b922987cf1259821358f3b462a8d05adf32adbdf0afd3a99dad066aab3b84b
3
+ size 98465062248
model-00006-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f29af9854d1d3cb678cf22322a22f56b86f61c0b11d7612fc52a3bd02e0a6977
3
+ size 98465062248
model-00007-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecc3076463fb6f0d1be98df68d7eb7eaf6f271b8eea1b8d6eec976b67f32fc6f
3
+ size 98465062248
model-00008-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58e418fd97e3076d2c936baa3fd694748f380458a7007adeefa4cb6574e005c5
3
+ size 98465062248
model-00009-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8696f78a45dcb1e0850f5ceb8a69893ee34b7882ff2ccefe6597da8aa4b1d879
3
+ size 98465062248
model-00010-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:79d82a903c854b0890d58e8a0065d37bc5bb74d56ac9bd7a72783d03ff4fdbef
3
+ size 98465062248
model-00011-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a743c8f9bb6cf14c4e4e27318df4722b3977e9d23d125ee88b00ef085ac93840
3
+ size 98465062248
model-00012-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8855321ca757aa9df1ea84b35da8300294e621bc964444342ecb7189d09e817f
3
+ size 98465062248
model-00013-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b215f3eb391ec44c8b61a0b75bdcd390d3703c8dec1458e05a07e3093f4a1202
3
+ size 98465062248
model-00014-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:67df94658eba7ed189a700533f6b0eb1413e230fe8dde0ae66acd4d419436b62
3
+ size 98465062248
model-00015-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b07d4cc22a97693665aae8343dfd52f8c21e6cab9f31ab4e58b105fbdec51859
3
+ size 98465062248
model-00016-of-00016.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49ea19923fa1761515b7c9a62013aa786bb4289fd9cc343c426f86a8515b9641
3
+ size 24313508504
model-dspark.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1ea6acb639ed5348618db4da5dcc1e4c4c973235f332df38f2d271a0bf051192
3
+ size 5064791810
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b657dadb90bf2855dcd931c78bf04f32874c2ece3a14ed1bd398eef7f8f23d98
3
+ size 12160261
tokenizer_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "bos_token": "[BOS]",
4
+ "clean_up_tokenization_spaces": false,
5
+ "eos_token": "<|endofturn|>",
6
+ "errors": "replace",
7
+ "is_local": true,
8
+ "local_files_only": false,
9
+ "model_input_names": [
10
+ "input_ids",
11
+ "attention_mask"
12
+ ],
13
+ "model_max_length": 1000000000000000019884624838656,
14
+ "pad_token": "[PAD]",
15
+ "padding_side": "right",
16
+ "split_special_tokens": false,
17
+ "tokenizer_class": "TokenizersBackend",
18
+ "unk_token": "[UNK]"
19
+ }