├── requirements.txt ├── requirements-dev.txt ├── .vscode └── extensions.json ├── streamlit_app.py ├── LICENSE └── README.md /requirements.txt: -------------------------------------------------------------------------------- 1 | pandas~=2.0.1 2 | streamlit~=1.22.0 -------------------------------------------------------------------------------- /requirements-dev.txt: -------------------------------------------------------------------------------- 1 | black 2 | flake 3 | isort 4 | mypy -------------------------------------------------------------------------------- /.vscode/extensions.json: -------------------------------------------------------------------------------- 1 | { 2 | "recommendations": [ 3 | "takumii.markdowntable" 4 | ] 5 | } -------------------------------------------------------------------------------- /streamlit_app.py: -------------------------------------------------------------------------------- 1 | import io 2 | import re 3 | from collections.abc import Iterable 4 | 5 | import pandas as pd 6 | import streamlit as st 7 | from pandas.api.types import is_bool_dtype, is_datetime64_any_dtype, is_numeric_dtype 8 | 9 | GITHUB_URL = "https://github.com/LudwigStumpp/llm-leaderboard" 10 | NON_BENCHMARK_COLS = ["Open?", "Publisher"] 11 | 12 | 13 | def extract_table_and_format_from_markdown_text(markdown_table: str) -> pd.DataFrame: 14 | """Extracts a table from a markdown text and formats it as a pandas DataFrame. 15 | 16 | Args: 17 | text (str): Markdown text containing a table. 18 | 19 | Returns: 20 | pd.DataFrame: Table as pandas DataFrame. 21 | """ 22 | df = ( 23 | pd.read_table(io.StringIO(markdown_table), sep="|", header=0, index_col=1) 24 | .dropna(axis=1, how="all") # drop empty columns 25 | .iloc[1:] # drop first row which is the "----" separator of the original markdown table 26 | .sort_index(ascending=True) 27 | .apply(lambda x: x.str.strip() if x.dtype == "object" else x) 28 | .replace("", float("NaN")) 29 | .astype(float, errors="ignore") 30 | ) 31 | 32 | # remove whitespace from column names and index 33 | df.columns = df.columns.str.strip() 34 | df.index = df.index.str.strip() 35 | df.index.name = df.index.name.strip() 36 | 37 | return df 38 | 39 | 40 | def extract_markdown_table_from_multiline(multiline: str, table_headline: str, next_headline_start: str = "#") -> str: 41 | """Extracts the markdown table from a multiline string. 42 | 43 | Args: 44 | multiline (str): content of README.md file. 45 | table_headline (str): Headline of the table in the README.md file. 46 | next_headline_start (str, optional): Start of the next headline. Defaults to "#". 47 | 48 | Returns: 49 | str: Markdown table. 50 | 51 | Raises: 52 | ValueError: If the table could not be found. 53 | """ 54 | # extract everything between the table headline and the next headline 55 | table = [] 56 | start = False 57 | for line in multiline.split("\n"): 58 | if line.startswith(table_headline): 59 | start = True 60 | elif line.startswith(next_headline_start): 61 | start = False 62 | elif start: 63 | table.append(line + "\n") 64 | 65 | if len(table) == 0: 66 | raise ValueError(f"Could not find table with headline '{table_headline}'") 67 | 68 | return "".join(table) 69 | 70 | 71 | def remove_markdown_links(text: str) -> str: 72 | """Modifies a markdown text to remove all markdown links. 73 | Example: [DISPLAY](LINK) to DISPLAY 74 | First find all markdown links with regex. 75 | Then replace them with: $1 76 | Args: 77 | text (str): Markdown text containing markdown links 78 | Returns: 79 | str: Markdown text without markdown links. 80 | """ 81 | 82 | # find all markdown links 83 | markdown_links = re.findall(r"\[([^\]]+)\]\(([^)]+)\)", text) 84 | 85 | # remove link keep display text 86 | for display, link in markdown_links: 87 | text = text.replace(f"[{display}]({link})", display) 88 | 89 | return text 90 | 91 | 92 | def filter_dataframe_by_row_and_columns(df: pd.DataFrame, ignore_columns: list[str] | None = None) -> pd.DataFrame: 93 | """ 94 | Filter dataframe by the rows and columns to display. 95 | 96 | This does not select based on the values in the dataframe, but rather on the index and columns. 97 | Modified from https://blog.streamlit.io/auto-generate-a-dataframe-filtering-ui-in-streamlit-with-filter_dataframe/ 98 | 99 | Args: 100 | df (pd.DataFrame): Original dataframe 101 | ignore_columns (list[str], optional): Columns to ignore. Defaults to None. 102 | 103 | Returns: 104 | pd.DataFrame: Filtered dataframe 105 | """ 106 | df = df.copy() 107 | 108 | if ignore_columns is None: 109 | ignore_columns = [] 110 | 111 | modification_container = st.container() 112 | 113 | with modification_container: 114 | to_filter_index = st.multiselect("Filter by model:", sorted(df.index)) 115 | if to_filter_index: 116 | df = pd.DataFrame(df.loc[to_filter_index]) 117 | 118 | to_filter_columns = st.multiselect( 119 | "Filter by benchmark:", sorted([c for c in df.columns if c not in ignore_columns]) 120 | ) 121 | if to_filter_columns: 122 | df = pd.DataFrame(df[ignore_columns + to_filter_columns]) 123 | 124 | return df 125 | 126 | 127 | def filter_dataframe_by_column_values(df: pd.DataFrame) -> pd.DataFrame: 128 | """ 129 | Filter dataframe by the values in the dataframe. 130 | 131 | Modified from https://blog.streamlit.io/auto-generate-a-dataframe-filtering-ui-in-streamlit-with-filter_dataframe/ 132 | 133 | Args: 134 | df (pd.DataFrame): Original dataframe 135 | 136 | Returns: 137 | pd.DataFrame: Filtered dataframe 138 | """ 139 | df = df.copy() 140 | 141 | modification_container = st.container() 142 | 143 | with modification_container: 144 | to_filter_columns = st.multiselect("Filter results on:", df.columns) 145 | left, right = st.columns((1, 20)) 146 | 147 | for column in to_filter_columns: 148 | if is_bool_dtype(df[column]): 149 | user_bool_input = right.checkbox(f"{column}", value=True) 150 | df = df[df[column] == user_bool_input] 151 | 152 | elif is_numeric_dtype(df[column]): 153 | _min = float(df[column].min()) 154 | _max = float(df[column].max()) 155 | 156 | if (_min != _max) and pd.notna(_min) and pd.notna(_max): 157 | step = 0.01 158 | user_num_input = right.slider( 159 | f"Values for {column}:", 160 | min_value=round(_min - step, 2), 161 | max_value=round(_max + step, 2), 162 | value=(_min, _max), 163 | step=step, 164 | ) 165 | df = df[df[column].between(*user_num_input)] 166 | 167 | elif is_datetime64_any_dtype(df[column]): 168 | user_date_input = right.date_input( 169 | f"Values for {column}:", 170 | value=( 171 | df[column].min(), 172 | df[column].max(), 173 | ), 174 | ) 175 | if isinstance(user_date_input, Iterable) and len(user_date_input) == 2: 176 | user_date_input_datetime = tuple(map(pd.to_datetime, user_date_input)) 177 | start_date, end_date = user_date_input_datetime 178 | df = df.loc[df[column].between(start_date, end_date)] 179 | 180 | else: 181 | selected_values = right.multiselect( 182 | f"Values for {column}:", 183 | sorted(df[column].unique()), 184 | ) 185 | 186 | if selected_values: 187 | df = df[df[column].isin(selected_values)] 188 | 189 | return df 190 | 191 | 192 | def setup_basic(): 193 | title = "🏆 LLM-Leaderboard" 194 | 195 | st.set_page_config( 196 | page_title=title, 197 | page_icon="🏆", 198 | layout="wide", 199 | ) 200 | st.title(title) 201 | 202 | st.markdown( 203 | "A joint community effort to create one central leaderboard for LLMs." 204 | f" Visit [llm-leaderboard]({GITHUB_URL}) to contribute. \n" 205 | 'We refer to a model being "open" if it can be locally deployed and used for commercial purposes.' 206 | ) 207 | 208 | 209 | def setup_leaderboard(readme: str): 210 | leaderboard_table = extract_markdown_table_from_multiline(readme, table_headline="## Leaderboard") 211 | leaderboard_table = remove_markdown_links(leaderboard_table) 212 | df_leaderboard = extract_table_and_format_from_markdown_text(leaderboard_table) 213 | df_leaderboard["Open?"] = df_leaderboard["Open?"].map({"yes": 1, "no": 0}).astype(bool) 214 | 215 | st.markdown("## Leaderboard") 216 | modify = st.checkbox("Add filters") 217 | clear_empty_entries = st.checkbox("Clear empty entries", value=True) 218 | 219 | if modify: 220 | df_leaderboard = filter_dataframe_by_row_and_columns(df_leaderboard, ignore_columns=NON_BENCHMARK_COLS) 221 | df_leaderboard = filter_dataframe_by_column_values(df_leaderboard) 222 | 223 | if clear_empty_entries: 224 | df_leaderboard = df_leaderboard.dropna(axis=1, how="all") 225 | benchmark_columns = [c for c in df_leaderboard.columns if df_leaderboard[c].dtype == float] 226 | rows_wo_any_benchmark = df_leaderboard[benchmark_columns].isna().all(axis=1) 227 | df_leaderboard = df_leaderboard[~rows_wo_any_benchmark] 228 | 229 | st.dataframe(df_leaderboard) 230 | 231 | st.download_button( 232 | "Download current selection as .html", 233 | df_leaderboard.to_html().encode("utf-8"), 234 | "leaderboard.html", 235 | "text/html", 236 | key="download-html", 237 | ) 238 | 239 | st.download_button( 240 | "Download current selection as .csv", 241 | df_leaderboard.to_csv().encode("utf-8"), 242 | "leaderboard.csv", 243 | "text/csv", 244 | key="download-csv", 245 | ) 246 | 247 | 248 | def setup_benchmarks(readme: str): 249 | benchmarks_table = extract_markdown_table_from_multiline(readme, table_headline="## Benchmarks") 250 | df_benchmarks = extract_table_and_format_from_markdown_text(benchmarks_table) 251 | 252 | st.markdown("## Covered Benchmarks") 253 | 254 | selected_benchmark = st.selectbox("Select a benchmark to learn more:", df_benchmarks.index.unique()) 255 | df_selected = df_benchmarks.loc[selected_benchmark] 256 | text = [ 257 | f"Name: {selected_benchmark}", 258 | ] 259 | for key in df_selected.keys(): 260 | text.append(f"{key}: {df_selected[key]} ") 261 | st.markdown(" \n".join(text)) 262 | 263 | 264 | def setup_sources(): 265 | st.markdown("## Sources") 266 | st.markdown( 267 | "The results of this leaderboard are collected from the individual papers and published results of the model " 268 | "authors. If you are interested in the sources of each individual reported model value, please visit the " 269 | f"[llm-leaderboard]({GITHUB_URL}) repository." 270 | ) 271 | st.markdown( 272 | """ 273 | Special thanks to the following pages: 274 | - [MosaicML - Model benchmarks](https://www.mosaicml.com/blog/mpt-7b) 275 | - [lmsys.org - Chatbot Arena benchmarks](https://lmsys.org/blog/2023-05-03-arena/) 276 | - [Papers With Code](https://paperswithcode.com/) 277 | - [Stanford HELM](https://crfm.stanford.edu/helm/latest/) 278 | - [HF Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) 279 | """ 280 | ) 281 | 282 | 283 | def setup_disclaimer(): 284 | st.markdown("## Disclaimer") 285 | st.markdown( 286 | "Above information may be wrong. If you want to use a published model for commercial use, please contact a " 287 | "lawyer." 288 | ) 289 | 290 | 291 | def setup_footer(): 292 | st.markdown( 293 | """ 294 | --- 295 | Made with ❤️ by the awesome open-source community from all over 🌍. 296 | """ 297 | ) 298 | 299 | 300 | def main(): 301 | setup_basic() 302 | 303 | with open("README.md", "r") as f: 304 | readme = f.read() 305 | 306 | setup_leaderboard(readme) 307 | setup_benchmarks(readme) 308 | setup_sources() 309 | setup_disclaimer() 310 | setup_footer() 311 | 312 | 313 | if __name__ == "__main__": 314 | main() 315 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | Apache License 2 | Version 2.0, January 2004 3 | http://www.apache.org/licenses/ 4 | 5 | TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION 6 | 7 | 1. Definitions. 8 | 9 | "License" shall mean the terms and conditions for use, reproduction, 10 | and distribution as defined by Sections 1 through 9 of this document. 11 | 12 | "Licensor" shall mean the copyright owner or entity authorized by 13 | the copyright owner that is granting the License. 14 | 15 | "Legal Entity" shall mean the union of the acting entity and all 16 | other entities that control, are controlled by, or are under common 17 | control with that entity. For the purposes of this definition, 18 | "control" means (i) the power, direct or indirect, to cause the 19 | direction or management of such entity, whether by contract or 20 | otherwise, or (ii) ownership of fifty percent (50%) or more of the 21 | outstanding shares, or (iii) beneficial ownership of such entity. 22 | 23 | "You" (or "Your") shall mean an individual or Legal Entity 24 | exercising permissions granted by this License. 25 | 26 | "Source" form shall mean the preferred form for making modifications, 27 | including but not limited to software source code, documentation 28 | source, and configuration files. 29 | 30 | "Object" form shall mean any form resulting from mechanical 31 | transformation or translation of a Source form, including but 32 | not limited to compiled object code, generated documentation, 33 | and conversions to other media types. 34 | 35 | "Work" shall mean the work of authorship, whether in Source or 36 | Object form, made available under the License, as indicated by a 37 | copyright notice that is included in or attached to the work 38 | (an example is provided in the Appendix below). 39 | 40 | "Derivative Works" shall mean any work, whether in Source or Object 41 | form, that is based on (or derived from) the Work and for which the 42 | editorial revisions, annotations, elaborations, or other modifications 43 | represent, as a whole, an original work of authorship. For the purposes 44 | of this License, Derivative Works shall not include works that remain 45 | separable from, or merely link (or bind by name) to the interfaces of, 46 | the Work and Derivative Works thereof. 47 | 48 | "Contribution" shall mean any work of authorship, including 49 | the original version of the Work and any modifications or additions 50 | to that Work or Derivative Works thereof, that is intentionally 51 | submitted to Licensor for inclusion in the Work by the copyright owner 52 | or by an individual or Legal Entity authorized to submit on behalf of 53 | the copyright owner. For the purposes of this definition, "submitted" 54 | means any form of electronic, verbal, or written communication sent 55 | to the Licensor or its representatives, including but not limited to 56 | communication on electronic mailing lists, source code control systems, 57 | and issue tracking systems that are managed by, or on behalf of, the 58 | Licensor for the purpose of discussing and improving the Work, but 59 | excluding communication that is conspicuously marked or otherwise 60 | designated in writing by the copyright owner as "Not a Contribution." 61 | 62 | "Contributor" shall mean Licensor and any individual or Legal Entity 63 | on behalf of whom a Contribution has been received by Licensor and 64 | subsequently incorporated within the Work. 65 | 66 | 2. Grant of Copyright License. Subject to the terms and conditions of 67 | this License, each Contributor hereby grants to You a perpetual, 68 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 69 | copyright license to reproduce, prepare Derivative Works of, 70 | publicly display, publicly perform, sublicense, and distribute the 71 | Work and such Derivative Works in Source or Object form. 72 | 73 | 3. Grant of Patent License. Subject to the terms and conditions of 74 | this License, each Contributor hereby grants to You a perpetual, 75 | worldwide, non-exclusive, no-charge, royalty-free, irrevocable 76 | (except as stated in this section) patent license to make, have made, 77 | use, offer to sell, sell, import, and otherwise transfer the Work, 78 | where such license applies only to those patent claims licensable 79 | by such Contributor that are necessarily infringed by their 80 | Contribution(s) alone or by combination of their Contribution(s) 81 | with the Work to which such Contribution(s) was submitted. If You 82 | institute patent litigation against any entity (including a 83 | cross-claim or counterclaim in a lawsuit) alleging that the Work 84 | or a Contribution incorporated within the Work constitutes direct 85 | or contributory patent infringement, then any patent licenses 86 | granted to You under this License for that Work shall terminate 87 | as of the date such litigation is filed. 88 | 89 | 4. Redistribution. You may reproduce and distribute copies of the 90 | Work or Derivative Works thereof in any medium, with or without 91 | modifications, and in Source or Object form, provided that You 92 | meet the following conditions: 93 | 94 | (a) You must give any other recipients of the Work or 95 | Derivative Works a copy of this License; and 96 | 97 | (b) You must cause any modified files to carry prominent notices 98 | stating that You changed the files; and 99 | 100 | (c) You must retain, in the Source form of any Derivative Works 101 | that You distribute, all copyright, patent, trademark, and 102 | attribution notices from the Source form of the Work, 103 | excluding those notices that do not pertain to any part of 104 | the Derivative Works; and 105 | 106 | (d) If the Work includes a "NOTICE" text file as part of its 107 | distribution, then any Derivative Works that You distribute must 108 | include a readable copy of the attribution notices contained 109 | within such NOTICE file, excluding those notices that do not 110 | pertain to any part of the Derivative Works, in at least one 111 | of the following places: within a NOTICE text file distributed 112 | as part of the Derivative Works; within the Source form or 113 | documentation, if provided along with the Derivative Works; or, 114 | within a display generated by the Derivative Works, if and 115 | wherever such third-party notices normally appear. The contents 116 | of the NOTICE file are for informational purposes only and 117 | do not modify the License. You may add Your own attribution 118 | notices within Derivative Works that You distribute, alongside 119 | or as an addendum to the NOTICE text from the Work, provided 120 | that such additional attribution notices cannot be construed 121 | as modifying the License. 122 | 123 | You may add Your own copyright statement to Your modifications and 124 | may provide additional or different license terms and conditions 125 | for use, reproduction, or distribution of Your modifications, or 126 | for any such Derivative Works as a whole, provided Your use, 127 | reproduction, and distribution of the Work otherwise complies with 128 | the conditions stated in this License. 129 | 130 | 5. Submission of Contributions. Unless You explicitly state otherwise, 131 | any Contribution intentionally submitted for inclusion in the Work 132 | by You to the Licensor shall be under the terms and conditions of 133 | this License, without any additional terms or conditions. 134 | Notwithstanding the above, nothing herein shall supersede or modify 135 | the terms of any separate license agreement you may have executed 136 | with Licensor regarding such Contributions. 137 | 138 | 6. Trademarks. This License does not grant permission to use the trade 139 | names, trademarks, service marks, or product names of the Licensor, 140 | except as required for reasonable and customary use in describing the 141 | origin of the Work and reproducing the content of the NOTICE file. 142 | 143 | 7. Disclaimer of Warranty. Unless required by applicable law or 144 | agreed to in writing, Licensor provides the Work (and each 145 | Contributor provides its Contributions) on an "AS IS" BASIS, 146 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or 147 | implied, including, without limitation, any warranties or conditions 148 | of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A 149 | PARTICULAR PURPOSE. You are solely responsible for determining the 150 | appropriateness of using or redistributing the Work and assume any 151 | risks associated with Your exercise of permissions under this License. 152 | 153 | 8. Limitation of Liability. In no event and under no legal theory, 154 | whether in tort (including negligence), contract, or otherwise, 155 | unless required by applicable law (such as deliberate and grossly 156 | negligent acts) or agreed to in writing, shall any Contributor be 157 | liable to You for damages, including any direct, indirect, special, 158 | incidental, or consequential damages of any character arising as a 159 | result of this License or out of the use or inability to use the 160 | Work (including but not limited to damages for loss of goodwill, 161 | work stoppage, computer failure or malfunction, or any and all 162 | other commercial damages or losses), even if such Contributor 163 | has been advised of the possibility of such damages. 164 | 165 | 9. Accepting Warranty or Additional Liability. While redistributing 166 | the Work or Derivative Works thereof, You may choose to offer, 167 | and charge a fee for, acceptance of support, warranty, indemnity, 168 | or other liability obligations and/or rights consistent with this 169 | License. However, in accepting such obligations, You may act only 170 | on Your own behalf and on Your sole responsibility, not on behalf 171 | of any other Contributor, and only if You agree to indemnify, 172 | defend, and hold each Contributor harmless for any liability 173 | incurred by, or claims asserted against, such Contributor by reason 174 | of your accepting any such warranty or additional liability. 175 | 176 | END OF TERMS AND CONDITIONS 177 | 178 | APPENDIX: How to apply the Apache License to your work. 179 | 180 | To apply the Apache License to your work, attach the following 181 | boilerplate notice, with the fields enclosed by brackets "[]" 182 | replaced with your own identifying information. (Don't include 183 | the brackets!) The text should be enclosed in the appropriate 184 | comment syntax for the file format. We also recommend that a 185 | file or class name and description of purpose be included on the 186 | same "printed page" as the copyright notice for easier 187 | identification within third-party archives. 188 | 189 | Copyright [yyyy] [name of copyright owner] 190 | 191 | Licensed under the Apache License, Version 2.0 (the "License"); 192 | you may not use this file except in compliance with the License. 193 | You may obtain a copy of the License at 194 | 195 | http://www.apache.org/licenses/LICENSE-2.0 196 | 197 | Unless required by applicable law or agreed to in writing, software 198 | distributed under the License is distributed on an "AS IS" BASIS, 199 | WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 200 | See the License for the specific language governing permissions and 201 | limitations under the License. 202 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # 🏆 LLM-Leaderboard 2 | 3 | A joint community effort to create one central leaderboard for LLMs. Contributions and corrections welcome!
4 | We refer to a model being "open" if it can be locally deployed and used for commercial purposes. 5 | 6 | ## Interactive Dashboard 7 | 8 | https://llm-leaderboard.streamlit.app/
9 | https://huggingface.co/spaces/ludwigstumpp/llm-leaderboard 10 | 11 | ## Leaderboard 12 | 13 | | Model Name | Publisher | Open? | Chatbot Arena Elo | HellaSwag (few-shot) | HellaSwag (zero-shot) | HellaSwag (one-shot) | HumanEval-Python (pass@1) | LAMBADA (zero-shot) | LAMBADA (one-shot) | MMLU (zero-shot) | MMLU (few-shot) | TriviaQA (zero-shot) | TriviaQA (one-shot) | WinoGrande (zero-shot) | WinoGrande (one-shot) | WinoGrande (few-shot) | 14 | | ----------------------------------------------------------------------------------------------------------- | ------------------- | ----- | ------------------------------------------------ | ------------------------------------------------------------------------- | ------------------------------------------------------------------ | --------------------------------------------------------------- | ------------------------------------------------------------------------------- | --------------------------------------------- | --------------------------------------------------------------- | ---------------------------------------------------------------------------------------- | ------------------------------------------------------------------------- | --------------------------------------------- | --------------------------------------------------------------- | ------------------------------------------------------------------ | --------------------------------------------------------------- | --------------------------------------------------------------- | 15 | | [alpaca-7b](https://crfm.stanford.edu/2023/03/13/alpaca.html) | Stanford | no | | | [0.739](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.661](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 16 | | [alpaca-13b](https://crfm.stanford.edu/2023/03/13/alpaca.html) | Stanford | no | [1008](https://lmsys.org/blog/2023-05-03-arena/) | | | | | | | | | | | | | | 17 | | [bloom-176b](https://huggingface.co/bigscience/bloom) | BigScience | yes | | [0.744](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | [0.155](https://huggingface.co/bigscience/bloom#results) | | | [0.299](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | | 18 | | [cerebras-gpt-7b](https://huggingface.co/cerebras/Cerebras-GPT-6.7B) | Cerebras | yes | | | [0.636](https://www.mosaicml.com/blog/mpt-7b) | | | [0.636](https://www.mosaicml.com/blog/mpt-7b) | | [0.259](https://www.mosaicml.com/blog/mpt-7b) | | [0.141](https://www.mosaicml.com/blog/mpt-7b) | | | | | 19 | | [cerebras-gpt-13b](https://huggingface.co/cerebras/Cerebras-GPT-13B) | Cerebras | yes | | | [0.635](https://www.mosaicml.com/blog/mpt-7b) | | | [0.635](https://www.mosaicml.com/blog/mpt-7b) | | [0.258](https://www.mosaicml.com/blog/mpt-7b) | | [0.146](https://www.mosaicml.com/blog/mpt-7b) | | | | | 20 | | [chatglm-6b](https://chatglm.cn/blog) | ChatGLM | yes | [985](https://lmsys.org/blog/2023-05-03-arena/) | | | | | | | | | | | | | | 21 | | [chinchilla-70b](https://arxiv.org/abs/2203.15556v1) | DeepMind | no | | | [0.808](https://arxiv.org/abs/2203.15556v1) | | | [0.774](https://arxiv.org/abs/2203.15556v1) | | | [0.675](https://arxiv.org/abs/2203.15556v1) | | | [0.749](https://arxiv.org/abs/2203.15556v1) | | | 22 | | [codex-12b / code-cushman-001](https://arxiv.org/abs/2107.03374) | OpenAI | no | | | | | [0.317](https://crfm.stanford.edu/helm/latest/?group=targeted_evaluations) | | | | | | | | | | 23 | | [codegen-16B-mono](https://huggingface.co/Salesforce/codegen-16B-mono) | Salesforce | yes | | | | | [0.293](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | | | | 24 | | [codegen-16B-multi](https://huggingface.co/Salesforce/codegen-16B-multi) | Salesforce | yes | | | | | [0.183](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | | | | 25 | | [codegx-13b](http://keg.cs.tsinghua.edu.cn/codegeex/) | Tsinghua University | no | | | | | [0.229](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | | | | 26 | | [dolly-v2-12b](https://huggingface.co/databricks/dolly-v2-12b) | Databricks | yes | [944](https://lmsys.org/blog/2023-05-03-arena/) | | [0.710](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.622](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 27 | | [eleuther-pythia-7b](https://huggingface.co/EleutherAI/pythia-6.9b) | EleutherAI | yes | | | [0.667](https://www.mosaicml.com/blog/mpt-7b) | | | [0.667](https://www.mosaicml.com/blog/mpt-7b) | | [0.265](https://www.mosaicml.com/blog/mpt-7b) | | [0.198](https://www.mosaicml.com/blog/mpt-7b) | | [0.661](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 28 | | [eleuther-pythia-12b](https://huggingface.co/EleutherAI/pythia-12b) | EleutherAI | yes | | | [0.704](https://www.mosaicml.com/blog/mpt-7b) | | | [0.704](https://www.mosaicml.com/blog/mpt-7b) | | [0.253](https://www.mosaicml.com/blog/mpt-7b) | | [0.233](https://www.mosaicml.com/blog/mpt-7b) | | [0.638](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 29 | | [falcon-7b](https://huggingface.co/tiiuae/falcon-40b) | TII | yes | | [0.781](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | | [0.350](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | 30 | | [falcon-40b](https://huggingface.co/tiiuae/falcon-40b) | TII | yes | | [0.853](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | | [0.527](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | 31 | | [fastchat-t5-3b](https://huggingface.co/lmsys/fastchat-t5-3b-v1.0) | Lmsys.org | yes | [951](https://lmsys.org/blog/2023-05-03-arena/) | | | | | | | | | | | | | | 32 | | [gal-120b](https://arxiv.org/abs/2211.09085v1) | Meta AI | no | | | | | | | | [0.526](https://paperswithcode.com/paper/galactica-a-large-language-model-for-science-1) | | | | | | | 33 | | [gpt-3-7b / curie](https://arxiv.org/abs/2005.14165) | OpenAI | no | | [0.682](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | | [0.243](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | 34 | | [gpt-3-175b / davinci](https://arxiv.org/abs/2005.14165) | OpenAI | no | | [0.793](https://arxiv.org/abs/2005.14165) | [0.789](https://arxiv.org/abs/2005.14165) | | | | | | [0.439](https://arxiv.org/abs/2005.14165) | | | [0.702](https://arxiv.org/abs/2005.14165v4) | | | 35 | | [gpt-3.5-175b / text-davinci-003](https://arxiv.org/abs/2303.08774v3) | OpenAI | no | | [0.822](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | [0.834](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | [0.481](https://arxiv.org/abs/2303.08774v3) | [0.762](https://arxiv.org/abs/2303.08774v3) | | | [0.569](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | [0.758](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | [0.816](https://arxiv.org/abs/2303.08774v3) | 36 | | [gpt-3.5-175b / code-davinci-002](https://platform.openai.com/docs/model-index-for-researchers) | OpenAI | no | | | | | [0.463](https://crfm.stanford.edu/helm/latest/?group=targeted_evaluations) | | | | | | | | | | 37 | | [gpt-4](https://arxiv.org/abs/2303.08774v3) | OpenAI | no | | [0.953](https://arxiv.org/abs/2303.08774v3) | | | [0.670](https://arxiv.org/abs/2303.08774v3) | | | | [0.864](https://arxiv.org/abs/2303.08774v3) | | | | | [0.875](https://arxiv.org/abs/2303.08774v3) | 38 | | [gpt4all-13b-snoozy](https://huggingface.co/nomic-ai/gpt4all-13b-snoozy) | Nomic AI | yes | | | [0.750](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.713](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 39 | | [gpt-neox-20b](https://huggingface.co/EleutherAI/gpt-neox-20b) | EleutherAI | yes | | [0.718](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | [0.719](https://www.mosaicml.com/blog/mpt-7b) | | | [0.719](https://www.mosaicml.com/blog/mpt-7b) | | [0.269](https://www.mosaicml.com/blog/mpt-7b) | [0.276](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | [0.347](https://www.mosaicml.com/blog/mpt-7b) | | | | | 40 | | [gpt-j-6b](https://huggingface.co/EleutherAI/gpt-j-6b) | EleutherAI | yes | | [0.663](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | [0.683](https://www.mosaicml.com/blog/mpt-7b) | | | [0.683](https://www.mosaicml.com/blog/mpt-7b) | | [0.261](https://www.mosaicml.com/blog/mpt-7b) | [0.249](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | [0.234](https://www.mosaicml.com/blog/mpt-7b) | | | | | 41 | | [koala-13b](https://bair.berkeley.edu/blog/2023/04/03/koala/) | Berkeley BAIR | no | [1082](https://lmsys.org/blog/2023-05-03-arena/) | | [0.726](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.688](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 42 | | [llama-7b](https://arxiv.org/abs/2302.13971) | Meta AI | no | | | [0.738](https://www.mosaicml.com/blog/mpt-7b) | | [0.105](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | [0.738](https://www.mosaicml.com/blog/mpt-7b) | | [0.302](https://www.mosaicml.com/blog/mpt-7b) | | [0.443](https://www.mosaicml.com/blog/mpt-7b) | | [0.701](https://arxiv.org/abs/2302.13971v1) | | | 43 | | [llama-13b](https://arxiv.org/abs/2302.13971) | Meta AI | no | [932](https://lmsys.org/blog/2023-05-03-arena/) | | [0.792](https://arxiv.org/abs/2302.13971) | | [0.158](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | [0.730](https://arxiv.org/abs/2302.13971v1) | | | 44 | | [llama-33b](https://arxiv.org/abs/2302.13971) | Meta AI | no | | | [0.828](https://arxiv.org/abs/2302.13971) | | [0.217](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | [0.760](https://arxiv.org/abs/2302.13971v1) | | | 45 | | [llama-65b](https://arxiv.org/abs/2302.13971) | Meta AI | no | | | [0.842](https://arxiv.org/abs/2302.13971) | | [0.237](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | [0.634](https://arxiv.org/abs/2302.13971v1) | | | [0.770](https://arxiv.org/abs/2302.13971v1) | | | 46 | | [llama-2-70b](https://arxiv.org/abs/2307.09288) | Meta AI | yes | | [0.873](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | | [0.698](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) | | | | | | 47 | | [mpt-7b](https://huggingface.co/mosaicml/mpt-7b) | MosaicML | yes | | | [0.761](https://www.mosaicml.com/blog/mpt-7b) | | | [0.702](https://www.mosaicml.com/blog/mpt-7b) | | [0.296](https://www.mosaicml.com/blog/mpt-7b) | | [0.343](https://www.mosaicml.com/blog/mpt-7b) | | | | | 48 | | [oasst-pythia-12b](https://huggingface.co/OpenAssistant/pythia-12b-pre-v8-12.5k-steps) | Open Assistant | yes | [1065](https://lmsys.org/blog/2023-05-03-arena/) | | [0.681](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.650](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 49 | | [opt-7b](https://huggingface.co/facebook/opt-6.7b) | Meta AI | no | | | [0.677](https://www.mosaicml.com/blog/mpt-7b) | | | [0.677](https://www.mosaicml.com/blog/mpt-7b) | | [0.251](https://www.mosaicml.com/blog/mpt-7b) | | [0.227](https://www.mosaicml.com/blog/mpt-7b) | | | | | 50 | | [opt-13b](https://huggingface.co/facebook/opt-13b) | Meta AI | no | | | [0.692](https://www.mosaicml.com/blog/mpt-7b) | | | [0.692](https://www.mosaicml.com/blog/mpt-7b) | | [0.257](https://www.mosaicml.com/blog/mpt-7b) | | [0.282](https://www.mosaicml.com/blog/mpt-7b) | | | | | 51 | | [opt-66b](https://huggingface.co/facebook/opt-66b) | Meta AI | no | | [0.745](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | | [0.276](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | 52 | | [opt-175b](https://ai.facebook.com/blog/democratizing-access-to-large-scale-language-models-with-opt-175b/) | Meta AI | no | | [0.791](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | | [0.318](https://crfm.stanford.edu/helm/latest/?group=core_scenarios) | | | | | | 53 | | [palm-62b](https://arxiv.org/abs/2204.02311v5) | Google Research | no | | | | | | | | | | | | [0.770](https://arxiv.org/abs/2204.02311) | | | 54 | | [palm-540b](https://arxiv.org/abs/2204.02311v5) | Google Research | no | | [0.838](https://arxiv.org/abs/2204.02311v5) | [0.834](https://arxiv.org/abs/2204.02311v5) | [0.836](https://ai.google/static/documents/palm2techreport.pdf) | [0.262](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | [0.779](https://arxiv.org/abs/2204.02311v5) | [0.818](https://ai.google/static/documents/palm2techreport.pdf) | | [0.693](https://arxiv.org/abs/2204.02311v5) | | [0.814](https://ai.google/static/documents/palm2techreport.pdf) | [0.811](https://arxiv.org/abs/2204.02311) | [0.837](https://ai.google/static/documents/palm2techreport.pdf) | [0.851](https://arxiv.org/abs/2204.02311) | 55 | | [palm-coder-540b](https://arxiv.org/abs/2204.02311) | Google Research | no | | | | | [0.359](https://ai.google/static/documents/palm2techreport.pdf) | | | | | | | | | | 56 | | [palm-2-s](https://ai.google/static/documents/palm2techreport.pdf) | Google Research | no | | | | [0.820](https://ai.google/static/documents/palm2techreport.pdf) | | | [0.807](https://ai.google/static/documents/palm2techreport.pdf) | | | | [0.752](https://ai.google/static/documents/palm2techreport.pdf) | | [0.779](https://ai.google/static/documents/palm2techreport.pdf) | | 57 | | [palm-2-s*](https://ai.google/static/documents/palm2techreport.pdf) | Google Research | no | | | | | [0.376](https://ai.google/static/documents/palm2techreport.pdf) | | | | | | | | | | 58 | | [palm-2-m](https://ai.google/static/documents/palm2techreport.pdf) | Google Research | no | | | | [0.840](https://ai.google/static/documents/palm2techreport.pdf) | | | [0.837](https://ai.google/static/documents/palm2techreport.pdf) | | | | [0.817](https://ai.google/static/documents/palm2techreport.pdf) | | [0.792](https://ai.google/static/documents/palm2techreport.pdf) | | 59 | | [palm-2-l](https://ai.google/static/documents/palm2techreport.pdf) | Google Research | no | | | | [0.868](https://ai.google/static/documents/palm2techreport.pdf) | | | [0.869](https://ai.google/static/documents/palm2techreport.pdf) | | | | [0.861](https://ai.google/static/documents/palm2techreport.pdf) | | [0.830](https://ai.google/static/documents/palm2techreport.pdf) | | 60 | | [palm-2-l-instruct](https://ai.google/static/documents/palm2techreport.pdf) | Google Research | no | | | | | | | | | | | | | | [0.909](https://ai.google/static/documents/palm2techreport.pdf) | 61 | | [replit-code-v1-3b](https://huggingface.co/replit/replit-code-v1-3b) | Replit | yes | | | | | [0.219](https://twitter.com/amasad/status/1651019556423598081/photo/2) | | | | | | | | | | 62 | | [stablelm-base-alpha-7b](https://huggingface.co/stabilityai/stablelm-base-alpha-7b) | Stability AI | yes | | | [0.412](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | [0.533](https://www.mosaicml.com/blog/mpt-7b) | | [0.251](https://www.mosaicml.com/blog/mpt-7b) | | [0.049](https://www.mosaicml.com/blog/mpt-7b) | | [0.501](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 63 | | [stablelm-tuned-alpha-7b](https://huggingface.co/stabilityai/stablelm-tuned-alpha-7b) | Stability AI | no | [858](https://lmsys.org/blog/2023-05-03-arena/) | | [0.536](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | | | | | | | [0.548](https://gpt4all.io/reports/GPT4All_Technical_Report_3.pdf) | | | 64 | | [starcoder-base-16b](https://huggingface.co/bigcode/starcoderbase) | BigCode | yes | | | | | [0.304](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | | | | 65 | | [starcoder-16b](https://huggingface.co/bigcode/starcoder) | BigCode | yes | | | | | [0.336](https://drive.google.com/file/d/1cN-b9GnWtHzQRoE7M7gAEyivY0kl4BYs/view) | | | | | | | | | | 66 | | [vicuna-13b](https://huggingface.co/lmsys/vicuna-13b-delta-v0) | Lmsys.org | no | [1169](https://lmsys.org/blog/2023-05-03-arena/) | | | | | | | | | | | | | | 67 | 68 | ## Benchmarks 69 | 70 | | Benchmark Name | Author | Link | Description | 71 | | ----------------- | ---------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | 72 | | Chatbot Arena Elo | LMSYS | https://lmsys.org/blog/2023-05-03-arena/ | "In this blog post, we introduce Chatbot Arena, an LLM benchmark platform featuring anonymous randomized battles in a crowdsourced manner. Chatbot Arena adopts the Elo rating system, which is a widely-used rating system in chess and other competitive games." (Source: https://lmsys.org/blog/2023-05-03-arena/) | 73 | | HellaSwag | Zellers et al. | https://arxiv.org/abs/1905.07830v1 | "HellaSwag is a challenge dataset for evaluating commonsense NLI that is specially hard for state-of-the-art models, though its questions are trivial for humans (>95% accuracy)." (Source: https://paperswithcode.com/dataset/hellaswag) | 74 | | HumanEval | Chen et al. | https://arxiv.org/abs/2107.03374v2 | "It used to measure functional correctness for synthesizing programs from docstrings. It consists of 164 original programming problems, assessing language comprehension, algorithms, and simple mathematics, with some comparable to simple software interview questions." (Source: https://paperswithcode.com/dataset/humaneval) | 75 | | LAMBADA | Paperno et al. | https://arxiv.org/abs/1606.06031 | "The LAMBADA evaluates the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not if they only see the last sentence preceding the target word. To succeed on LAMBADA, computational models cannot simply rely on local context, but must be able to keep track of information in the broader discourse." (Source: https://huggingface.co/datasets/lambada) | 76 | | MMLU | Hendrycks et al. | https://github.com/hendrycks/test | "The benchmark covers 57 subjects across STEM, the humanities, the social sciences, and more. It ranges in difficulty from an elementary level to an advanced professional level, and it tests both world knowledge and problem solving ability. Subjects range from traditional areas, such as mathematics and history, to more specialized areas like law and ethics. The granularity and breadth of the subjects makes the benchmark ideal for identifying a model’s blind spots." (Source: "https://paperswithcode.com/dataset/mmlu") | 77 | | TriviaQA | Joshi et al. | https://arxiv.org/abs/1705.03551v2 | "We present TriviaQA, a challenging reading comprehension dataset containing over 650K question-answer-evidence triples. TriviaQA includes 95K question-answer pairs authored by trivia enthusiasts and independently gathered evidence documents, six per question on average, that provide high quality distant supervision for answering the questions." (Source: https://arxiv.org/abs/1705.03551v2) | 78 | | WinoGrande | Sakaguchi et al. | https://arxiv.org/abs/1907.10641v2 | "A large-scale dataset of 44k [expert-crafted pronoun resolution] problems, inspired by the original WSC design, but adjusted to improve both the scale and the hardness of the dataset." (Source: https://arxiv.org/abs/1907.10641v2) | 79 | 80 | ## How to Contribute 81 | 82 | We are always happy for contributions! You can contribute by the following: 83 | 84 | - table work (don't forget the links): 85 | - filling missing entries 86 | - adding a new model as a new row to the leaderboard. Please keep alphabetic order. 87 | - adding a new benchmark as a new column in the leaderboard and add the benchmark to the benchmarks table. Please keep alphabetic order. 88 | - code work: 89 | - improving the existing code 90 | - requesting and implementing new features 91 | 92 | ## Future Ideas 93 | 94 | - (TBD) add model year 95 | - (TBD) add model details: 96 | - #params 97 | - #tokens seen during training 98 | - length context window 99 | - architecture type (transformer-decoder, transformer-encoder, transformer-encoder-decoder, ...) 100 | 101 | ## More Open LLMs 102 | 103 | If you are interested in an overview about open llms for commercial use and finetuning, check out the [open-llms](https://github.com/eugeneyan/open-llms) repository. 104 | 105 | ## Sources 106 | 107 | The results of this leaderboard are collected from the individual papers and published results of the model authors. For each reported value, the source is added as a link. 108 | 109 | Special thanks to the following pages: 110 | - [MosaicML - Model benchmarks](https://www.mosaicml.com/blog/mpt-7b) 111 | - [lmsys.org - Chatbot Arena benchmarks](https://lmsys.org/blog/2023-05-03-arena/) 112 | - [Papers With Code](https://paperswithcode.com/) 113 | - [Stanford HELM](https://crfm.stanford.edu/helm/latest/) 114 | - [HF Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard) 115 | 116 | ## Disclaimer 117 | 118 | Above information may be wrong. If you want to use a published model for commercial use, please contact a lawyer. --------------------------------------------------------------------------------