Qwen1.5 110B Chat

Name: Qwen1.5 110B Chat
Author: Alibaba

Alibaba · 2024-04-25 · 111.2B parameters

Qwen1.5-110B-Chat is Alibaba's 110B instruction-tuned chat model in the Qwen1.5 family, released April 2024. It is a dense transformer model fine-tuned for conversational tasks and was the largest open-weight model in the Qwen1.5 series at release.

Benchmark scores

Benchmark	Score
Arena Text — Business, Management & Financial Ops (No Style Control)	1140
Arena Text — Chinese (No Style Control)	1206
Arena Text — Creative Writing (No Style Control)	1148
Arena Text — English (No Style Control)	1198
Arena Text — Entertainment, Sports & Media (No Style Control)	1138
Arena Text — Exclude Ties (No Style Control)	1044
Arena Text — Exclude Ties (Style Controlled)	1129
Arena Text — Expert (No Style Control)	1144
Arena Text — French (No Style Control)	1151
Arena Text — French (Style Controlled)	1226
Arena Text — German (No Style Control)	1123
Arena Text — Hard Prompts (No Style Control)	1168
Arena Text — Hard Prompts English (No Style Control)	1185
Arena Text — Instruction Following (No Style Control)	1158
Arena Text — Japanese (No Style Control)	1074
Arena Text — Korean (No Style Control)	1044
Arena Text — Legal & Government (No Style Control)	1188
Arena Text — Life, Physical & Social Science (No Style Control)	1166
Arena Text — Longer Query (No Style Control)	1157
Arena Text — Medicine & Healthcare (No Style Control)	1146
Arena Text — Multi-turn (No Style Control)	1160
Arena Text — Non-English (No Style Control)	1142
Arena Text — Non-English (Style Controlled)	1202
Arena Text — Occupational Mathematical (No Style Control)	1190
Arena Text — Russian (No Style Control)	1118
Arena Text — Software & IT Services (No Style Control)	1174
Arena Text — Spanish (No Style Control)	1142
Arena Text — Writing, Literature & Language (No Style Control)	1170
BBH (Open LLM Leaderboard v2)	61.8
BigCodeBench Complete	44.4