Chatbot Arena

This leaderboard is based on the following benchmarks.

• Chatbot Arena - a crowdsourced, randomized battle platform for large language models (LLMs). We use 2.8M+ user votes to compute Elo ratings.
• MMLU - a test to measure a model’s multitask accuracy on 57 tasks.
• Arena-Hard-Auto - an automatic evaluation tool for instruction-tuned LLMs.

| Vote | Blog | GitHub | Paper | Dataset | Twitter | Discord |

Best Open LM

Model	Arena Elo	MMLU	License
DeepSeek-V3-0324	1370	88.5	MIT
DeepSeek-R1	1359	90.8	MIT
Gemma-3-27B-it	1342		Gemma
QwQ-32B	1315		Apache 2.0
Llama-3.3-Nemotron-Super-49B-v1	1296	86	Nvidia

Full Leaderboard

Model	Arena Elo	Coding	Vision	Arena Hard	MMLU	Votes	Organization	License
🥇 Gemini-2.5-Pro-Exp-03-25	1437	1423	1315			7431	Google	Proprietary
🥇 ChatGPT-4o-latest (2025-03-26)	1406	1416	1304			6612	OpenAI	Proprietary
🥇 Grok-3-Preview-02-24	1402	1409			92.7	13919	xAI	Proprietary
🥇 GPT-4.5-Preview	1397	1402	1253			13443	OpenAI	Proprietary
🥈 Gemini-2.0-Pro-Exp-02-05	1380	1379	1240			20136	Google	Proprietary
🥈 Gemini-2.0-Flash-Thinking-Exp-01-21	1380	1365	1279			25266	Google	Proprietary
🥈 DeepSeek-V3-0324	1370	1394			88.5	4721	DeepSeek	MIT
🥈 DeepSeek-R1	1359	1368			90.8	15098	DeepSeek	MIT
🥈 Gemini-2.0-Flash-Exp	1355	1353	1257			22518	Google	Proprietary
🥈 o1-2024-12-17	1350	1358	1227	90.4	91.8	27831	OpenAI	Proprietary
🥈 Gemma-3-27B-it	1342	1312				9147	Google	Gemma
🥈 Qwen2.5-Max	1340	1345				19995	Alibaba	Proprietary
🥈 o3-mini-high	1325	1364				16889	OpenAI	Proprietary
🥉 DeepSeek-V3	1318	1320		85.5	88.5	22843	DeepSeek	DeepSeek
🥉 QwQ-32B	1315	1323				6729	Alibaba	Apache 2.0
🥉 Qwen-Plus-0125	1310	1320				6058	Alibaba	Proprietary
🥉 Gemini-2.0-Flash-Lite	1310	1318	1156			20990	Google	Proprietary
🥉 GLM-4-Plus-0111	1310	1290				6032	Zhipu	Proprietary
🥉 o3-mini	1305	1348				23693	OpenAI	Proprietary
🥉 Command A (03-2025)	1305	1314				6380	Cohere	CC-BY-NC-4.0
🥉 Step-2-16K-Exp	1305	1295				5126	StepFun	Proprietary
🥉 o1-mini	1304	1353		92		54967	OpenAI	Proprietary
🥉 Claude 3.7 Sonnet (thinking-32k)	1303	1336				7809	Anthropic	Proprietary
🥉 Hunyuan-TurboS-20250226	1302	1327				2456	Tencent	Proprietary
🥉 Gemini-1.5-Pro-002	1302	1291	1222			58651	Google	Proprietary
🥉 Hunyuan-Turbo-0110	1296	1315				2509	Tencent	Proprietary
🥉 Llama-3.3-Nemotron-Super-49B-v1	1296	1301		88.3	86	2371	Nvidia	Nvidia
🥉 Claude 3.7 Sonnet	1295	1332	1214			13107	Anthropic	Proprietary
🥉 Grok-2-08-13	1288	1282			87.5	67092	xAI	Proprietary
🥉 Yi-Lightning	1287	1303		81.5		28971	01 AI	Proprietary
🥉 GPT-4o-2024-05-13	1285	1293	1206	79.21	88.7	117768	OpenAI	Proprietary
🥉 Claude 3.5 Sonnet (20241022)	1283	1326	1182	85.2	88.7	64098	Anthropic	Proprietary
Deepseek-v2.5-1210	1279	1297				7251	DeepSeek	DeepSeek
Athene-v2-Chat-72B	1275	1300		85		26094	NexusFlow	NexusFlow
Llama-4-Maverick-17B-128E-Instruct	1273	1299				2662	Meta	Llama 4
Hunyuan-Large-2025-02-10	1272	1292				3860	Tencent	Proprietary
GPT-4o-mini-2024-07-18	1272	1283	1124	74.94	82	70818	OpenAI	Proprietary
Gemini-1.5-Flash-002	1271	1254	1205			37026	Google	Proprietary
Llama-3.1-405B-Instruct-bf16	1269	1280			88.6	42624	Meta	Llama 3.1
Llama-3.1-Nemotron-70B-Instruct	1269	1271		84.9		7577	Nvidia	Llama 3.1
Llama-3.1-405B-Instruct-fp8	1267	1276		69.3	88.6	63061	Meta	Llama 3.1
Grok-2-Mini-08-13	1266	1262				55449	xAI	Proprietary
Yi-Lightning-lite	1264	1267				17073	01 AI	Proprietary
Hunyuan-Standard-2025-02-10	1260	1268				4014	Tencent	Proprietary
Qwen2.5-72B-Instruct	1257	1283		78		41535	Alibaba	Qwen
Llama-3.3-70B-Instruct	1257	1258				37031	Meta	Llama-3.3
GPT-4-Turbo-2024-04-09	1256	1263	1151	82.63		102160	OpenAI	Proprietary
Mistral-Large-2407	1251	1269		70.42		48219	Mistral	Mistral Research
GPT-4-1106-preview	1250	1253				103753	OpenAI	Proprietary
Athene-70B	1250	1253		77.6		20585	NexusFlow	CC-BY-NC-4.0
Mistral-Large-2411	1248	1265				28444	Mistral	MRL
Llama-3.1-70B-Instruct	1248	1251		55.73	86	58666	Meta	Llama 3.1
Claude 3 Opus	1247	1250	1076	60.36	86.8	202710	Anthropic	Proprietary
Amazon Nova Pro 1.0	1245	1261	1044			23635	Amazon	Proprietary
GPT-4-0125-preview	1245	1243		77.96		97077	OpenAI	Proprietary
Llama-3.1-Tulu-3-70B	1244	1233				3015	Ai2	Llama 3.1
Yi-Large-preview	1240	1245		71.48		51646	01 AI	Proprietary
Claude 3.5 Haiku (20241022)	1237	1264				32053	Anthropic	Propretary
Reka-Core-20240904	1235	1221				7938	Reka AI	Proprietary
Reka-Core-20240722	1230	1208				13286	Reka AI	Proprietary
Qwen-Plus-0828	1227	1245				14625	Alibaba	Proprietary
Gemini-1.5-Flash-001	1227	1232	1072	49.61	78.9	65668	Google	Proprietary
Jamba-1.5-Large	1221	1227			81.2	9127	AI21 Labs	Jamba Open
Deepseek-v2-API-0628	1220	1242				19505	DeepSeek AI	DeepSeek
Gemma-2-27B-it	1220	1209		57.51		79528	Google	Gemma license
Qwen2.5-Coder-32B-Instruct	1217	1261				5730	Alibaba	Apache 2.0
Amazon Nova Lite 1.0	1217	1235	1061			20655	Amazon	Proprietary
Mistral-Small-24B-Instruct-2501	1216	1230				13940	Mistral	Apache 2.0
Gemma-2-9B-it-SimPO	1216	1196				10552	Princeton	MIT
Command R+ (08-2024)	1215	1181				10541	Cohere	CC-BY-NC-4.0
Deepseek-Coder-v2-0724	1214	1266		62.3		11728	DeepSeek	Proprietary
Yi-Large	1212	1220		63.7		16628	01 AI	Proprietary
Gemini-1.5-Flash-8B-001	1212	1208	1106			37697	Google	Proprietary
Llama-3.1-Nemotron-51B-Instruct	1211	1211				3885	Nvidia	Llama 3.1
Nemotron-4-340B-Instruct	1209	1198				20613	Nvidia	NVIDIA Open Model
Aya-Expanse-32B	1209	1193				28751	Cohere	CC-BY-NC-4.0
GLM-4-0520	1206	1216		63.84		10221	Zhipu AI	Proprietary
Llama-3-70B-Instruct	1206	1200		46.57	82	163660	Meta	Llama 3
Reka-Flash-20240904	1205	1191				8138	Reka AI	Proprietary
Gemini-1.5-Flash-8B-Exp-0827	1205	1189	1112			25344	Google	Proprietary
Phi-4	1204	1221				24508	Microsoft	MIT
Claude 3 Sonnet	1201	1213	1048	46.8	79	113056	Anthropic	Proprietary
Reka-Flash-20240722	1201	1187				13729	Reka AI	Proprietary
Reka-Core-20240501	1199	1190	1015		83.2	62566	Reka AI	Proprietary
Amazon Nova Micro 1.0	1198	1210				20666	Amazon	Proprietary
Gemma-2-9B-it	1192	1173				57197	Google	Gemma license
Command R+ (04-2024)	1190	1164		33.07		80856	Cohere	CC-BY-NC-4.0
Hunyuan-Standard-256K	1189	1227				2901	Tencent	Proprietary
Qwen2-72B-Instruct	1187	1187		46.86	84.2	38884	Alibaba	Qianwen LICENSE
GPT-4-0314	1186	1195		50	86.4	55980	OpenAI	Proprietary
Llama-3.1-Tulu-3-8B	1185	1179				3075	Ai2	Llama 3.1
GLM-4-0116	1183	1191		55.72		7576	Zhipu AI	Proprietary
Qwen-Max-0428	1183	1189				25694	Alibaba	Proprietary
Ministral-8B-2410	1182	1201				5111	Mistral	MRL
Aya-Expanse-8B	1180	1165				10391	Cohere	CC-BY-NC-4.0
Claude 3 Haiku	1179	1189	1000	41.47	75.2	122305	Anthropic	Proprietary
Command R (08-2024)	1179	1161				10849	Cohere	CC-BY-NC-4.0
DeepSeek-Coder-V2-Instruct	1178	1239				15753	DeepSeek AI	DeepSeek License
Llama-3.1-8B-Instruct	1176	1186		21.34	73	52599	Meta	Llama 3.1
Jamba-1.5-Mini	1176	1181			69.7	9269	AI21 Labs	Jamba Open
Reka-Flash-Preview-20240611	1165	1155	1024			20430	Reka AI	Proprietary
GPT-4-0613	1163	1167		37.9		91639	OpenAI	Proprietary
Qwen1.5-110B-Chat	1161	1175			80.4	27441	Alibaba	Qianwen LICENSE
Mistral-Large-2402	1157	1170		37.71	81.2	64915	Mistral	Proprietary
Yi-1.5-34B-Chat	1157	1162			76.8	25136	01 AI	Apache-2.0
Reka-Flash-21B-online	1156	1147				16031	Reka AI	Proprietary
QwQ-32B-Preview	1153	1147				3413	Alibaba	Apache 2.0
Llama-3-8B-Instruct	1152	1146		20.56	68.4	109102	Meta	Llama 3
InternLM2.5-20B-chat	1149	1158				10595	InternLM	Other
Claude-1	1149	1136			77	21151	Anthropic	Proprietary
Command R (04-2024)	1149	1123		17.02		56382	Cohere	CC-BY-NC-4.0
Mistral Medium	1148	1152		31.9	75.3	35559	Mistral	Proprietary
Qwen1.5-72B-Chat	1147	1160		36.12	77.5	40669	Alibaba	Qianwen LICENSE
Mixtral-8x22b-Instruct-v0.1	1147	1153		36.36	77.8	53767	Mistral	Apache 2.0
Reka-Flash-21B	1147	1141			73.5	25813	Reka AI	Proprietary
Gemma-2-2b-it	1144	1107			51.3	48906	Google	Gemma license
Granite-3.1-8B-Instruct	1142	1173				3293	IBM	Apache 2.0
Claude-2.0	1132	1135		23.99	78.5	12763	Anthropic	Proprietary
Gemini-1.0-Pro-001	1131	1103			71.8	18802	Google	Proprietary
Zephyr-ORPO-141b-A35b-v0.1	1127	1124				4860	HuggingFace	Apache 2.0
Qwen1.5-32B-Chat	1125	1149			73.4	22762	Alibaba	Qianwen LICENSE
Mistral-Next	1124	1132		27.37		12376	Mistral	Proprietary
Phi-3-Medium-4k-Instruct	1123	1125		33.37	78	26111	Microsoft	MIT
Granite-3.1-2B-Instruct	1119	1147				3382	IBM	Apache 2.0
Starling-LM-7B-beta	1119	1129		23.01		16670	Nexusflow	Apache-2.0
Claude-2.1	1118	1132		22.77		37695	Anthropic	Proprietary
GPT-3.5-Turbo-0613	1117	1135		24.82		38958	OpenAI	Proprietary
Mixtral-8x7B-Instruct-v0.1	1114	1114		23.4	70.6	76138	Mistral	Apache 2.0
Claude-Instant-1	1111	1109			73.4	20625	Anthropic	Proprietary
Yi-34B-Chat	1111	1106		23.15	73.5	15919	01 AI	Yi License
Gemini Pro	1111	1091		17.8	71.8	6557	Google	Proprietary
Qwen1.5-14B-Chat	1109	1126			67.6	18691	Alibaba	Qianwen LICENSE
GPT-3.5-Turbo-0314	1107	1115		18.05	70	5639	OpenAI	Proprietary
GPT-3.5-Turbo-0125	1106	1124		23.34		68869	OpenAI	Proprietary
WizardLM-70B-v1.0	1106	1071			63.7	8382	Microsoft	Llama 2
DBRX-Instruct-Preview	1103	1118		24.63	73.7	33733	Databricks	DBRX LICENSE
Llama-3.2-3B-Instruct	1103	1080				8400	Meta	Llama 3.2
Phi-3-Small-8k-Instruct	1102	1107		29.77	75.7	18472	Microsoft	MIT
Tulu-2-DPO-70B	1099	1093		14.99		6659	AllenAI/UW	AI2 ImpACT Low-risk
Granite-3.0-8B-Instruct	1093	1097				7000	IBM	Apache 2.0
Llama-2-70B-chat	1093	1072		11.55	63	39608	Meta	Llama 2
OpenChat-3.5-0106	1091	1102			65.8	12990	OpenChat	Apache-2.0
Vicuna-33B	1091	1067		8.63	59.2	22945	LMSYS	Non-commercial
Snowflake Arctic Instruct	1090	1077		17.61	67.3	34184	Snowflake	Apache 2.0
Starling-LM-7B-alpha	1088	1080		12.8	63.9	10416	UC Berkeley	CC-BY-NC-4.0
Gemma-1.1-7B-it	1084	1084		12.09	64.3	25086	Google	Gemma license
Nous-Hermes-2-Mixtral-8x7B-DPO	1084	1079				3838	NousResearch	Apache-2.0
NV-Llama2-70B-SteerLM-Chat	1080	1023			68.5	3635	Nvidia	Llama 2
pplx-70B-online	1078	1028				6895	Perplexity AI	Proprietary
DeepSeek-LLM-67B-Chat	1077	1079			71.3	4987	DeepSeek AI	DeepSeek License
OpenChat-3.5	1076	1054			64.3	8107	OpenChat	Apache-2.0
Granite-3.0-2B-Instruct	1074	1088				7187	IBM	Apache 2.0
OpenHermes-2.5-Mistral-7B	1074	1058				5088	NousResearch	Apache-2.0
Mistral-7B-Instruct-v0.2	1072	1074		12.57		20065	Mistral	Apache-2.0
Qwen1.5-7B-Chat	1070	1089			61	4871	Alibaba	Qianwen LICENSE
Phi-3-Mini-4K-Instruct-June-24	1070	1082			70.9	12808	Microsoft	MIT
GPT-3.5-Turbo-1106	1068	1095		18.87		17032	OpenAI	Proprietary
Phi-3-Mini-4k-Instruct	1066	1086			68.8	21090	Microsoft	MIT
Llama-2-13b-chat	1063	1051			53.6	19719	Meta	Llama 2
SOLAR-10.7B-Instruct-v1.0	1062	1047			66.2	4287	Upstage AI	CC-BY-NC-4.0
Dolphin-2.2.1-Mistral-7B	1062	1025				1713	Cognitive Computations	Apache-2.0
WizardLM-13b-v1.2	1059	1026			52.7	7175	Microsoft	Llama 2
Llama-3.2-1B-Instruct	1054	1046				8519	Meta	Llama 3.2
Qwen2.5-VL-32B-Instruct			1218				Alibaba	Apache 2.0
Step-1o-Vision-32k (highres)			1187				StepFun	Proprietary
Qwen2.5-VL-72B-Instruct			1172				Alibaba	Qwen
Pixtral-Large-2411			1154				Mistral	MRL
Qwen-VL-Max-1119			1128				Alibaba	Proprietary
Step-1V-32K			1111				StepFun	Proprietary
Qwen2-VL-72b-Instruct			1110				Alibaba	Qwen
Molmo-72B-0924			1076				AI2	Apache 2.0
Pixtral-12B-2409			1072				Mistral	Apache 2.0
Llama-3.2-90B-Vision-Instruct			1069				Meta	Llama 3.2
Aya-Vision-8B			1069				Cohere	CC-BY-NC-4.0
InternVL2-26B			1067				OpenGVLab	MIT
Hunyuan-Standard-Vision-2024-12-31			1066				Tencent	Proprietary
Aya-Vision-32B			1057				Cohere	CC-BY-NC-4.0
Qwen2-VL-7B-Instruct			1054				Aliaba	Apache 2.0
Yi-Vision			1045				01 AI	Proprietary
Llama-3.2-11B-Vision-Instruct			1032				Meta	Llama 3.2

If you want to see more models, please help us add them.

💻 Code: The Arena Elo ratings are computed by this notebook. The MT-bench scores (single-answer grading on a scale of 10) are computed by fastchat.llm_judge. The MMLU scores are computed by InstructEval. Higher values are better for all benchmarks. Empty cells mean not available. The latest and detailed leaderboard is here.

More Statistics for Chatbot Arena

🔗 Arena Statistics

Transition from online Elo rating system to Bradley-Terry model

We adopted the Elo rating system for ranking models since the launch of the Arena. It has been useful to transform pairwise human preference to Elo ratings that serve as a predictor of winrate between models. Specifically, if player A has a rating of R_A and player B a rating of R_B, the probability of player A winning is

ELO rating has been used to rank chess players by the international community for over 60 years. Standard Elo rating systems assume a player’s performance changes overtime. So an online algorithm is needed to capture such dynamics, meaning recent games should weigh more than older games. Specifically, after each game, a player’s rating is updated according to the difference between predicted outcome and actual outcome.

This algorithm has two distinct features:

1. It can be computed asynchronously by players around the world.
2. It allows for players performance to change dynamically – it does not assume a fixed unknown value for the players rating.

This ability to adapt is determined by the parameter K which controls the magnitude of rating changes that can affect the overall result. A larger K essentially put more weight on the recent games, which may make sense for new players whose performance improves quickly. However as players become more senior and their performance “converges” then a smaller value of K is more appropriate. As a result, USCF adopted K based on the number of games and tournaments completed by the player (reference). That is, the Elo rating of a senior player changes slower than a new player.

When we launched the Arena, we noticed considerable variability in the ratings using the classic online algorithm. We tried to tune the K to be sufficiently stable while also allowing new models to move up quickly in the leaderboard. We ultimately decided to adopt a bootstrap-like technique to shuffle the data and sample Elo scores from 1000 permutations of the online plays. You can find the details in this notebook. This provided consistent stable scores and allowed us to incorporate new models quickly. This is also observed in a recent work by Cohere. However, we used the same samples to estimate confidence intervals which were therefore too wide (effectively CI’s for the original online Elo estimates).

In the context of LLM ranking, there are two important differences from the classic Elo chess ranking system. First, we have access to the entire history of all games for all models and so we don’t need a decentralized algorithm. Second, most models are static (we have access to the weights) and so we don’t expect their performance to change. However, it is worth noting that the hosted proprietary models may not be static and their behavior can change without notice. We try our best to pin specific model API versions if possible.

To improve the quality of our rankings and their confidence estimates, we are adopting another widely used rating system called the Bradley–Terry (BT) model. This model actually is the maximum likelihood (MLE) estimate of the underlying Elo model assuming a fixed but unknown pairwise win-rate. Similar to Elo rating, BT model is also based on pairwise comparison to derive ratings of players to estimate win rate between each other. The core difference between BT model vs the online Elo system is the assumption that player’s performance does not change (i.e., game order does not matter) and the computation takes place in a centralized fashion.