From votes to a leaderboard

This ranking is calculated from votes collected through blind comparisons, using the Bradley-Terry statistical model. It reflects subjective preferences and should be interpreted with caution: it measures neither factual accuracy nor overall model performance.

Total models: 121

Total votes: 268,000

Updated on 11/4/2026

Show ranks

Disabled

Tooltip Shows each model's exact position, from 1 to last. Two neighbouring ranks can almost never be told apart: the gap between 4th and 7th usually comes down to which votes happened to be cast. The classes say what the data supports.

Download data

Search for a model Search

From votes to a leaderboard

Rank Tooltip Rank assigned based on Bradley-Terry satisfaction score Sort by Rank	Model Sort by Model	BT score of satisfaction Tooltip Estimated statistical score based on the Bradley-Terry model, reflecting the probability that one model is preferred over another. This score is calculated from all user votes and reactions. For more information, visit the methodology tab. Sort by BT score of satisfaction	Confidence (±) Tooltip Interval indicating the reliability of the rank: the narrower the interval, the more reliable the rank estimate. There is a 95% chance that the model's true rank is within this range. Sort by Confidence (±)	Total votes Sort by Total votes	Average energy (per 1000 tokens) Tooltip Measured in watt-hours, energy consumption represents the electricity used by the model to process a request and generate the corresponding response. Model energy consumption depends on the model's size and architecture. We have chosen to display proprietary models for which we do not have transparent information about their size and architecture in gray as "unanalyzed" (N/A). Sort by Average energy (per 1000 tokens)	Size (parameters) Tooltip Model size in billions of parameters, categorized into five classes. For proprietary models, this size is not reported. Sort by Size (parameters)	Architecture Tooltip The architecture of an LLM model refers to the design principles that define how the components of a neural network are arranged and interact to transform input data into predictive outputs, including the activation mode of parameters (dense vs. sparse), the specialization of components, and the information processing mechanisms (transformers, convolutional networks, hybrid architectures). Sort by Architecture	Release date Sort by Release date	Organization Sort by Organization	Licence Sort by Licence
I Rank I, #1 to 3	claude-opus-5.5	1190	-2/+0	639	N / A	XL - (estimate)	N/A	9/26	Anthropic	Proprietary
I Rank I, #1 to 3	muse-spark-1-2	1178	-1/+1	886	N / A	XL - (estimate)	N/A	8/26	Meta	Proprietary
I Rank I, #1 to 3	muse-spark-1-3	1166	-1/+2	1002	N / A	XL - (estimate)	N/A	9/26	Meta	Proprietary
II Rank II, #4 to 17	gpt-5.3	1134	-5/+0	1176	N / A	L - (estimate)	N/A	3/26	OpenAI	Proprietary
II Rank II, #4 to 17	gemini-3-flash-preview	1132	-4/+1	3121	N / A	XL - (estimate)	N/A	12/25	Google	Proprietary
II Rank II, #4 to 17	claude-sonnet-5-5	1126	-6/+2	833	N / A	XL - (estimate)	N/A	9/26	Anthropic	Proprietary
II Rank II, #4 to 17	openai/gpt-6.1-sol	1121	-11/+3	763	N / A	XL - (estimate)	N/A	9/26	OpenAI	Proprietary
II Rank II, #4 to 17	gemini-3.1-flash-lite-preview	1119	-5/+4	2071	N / A	L - (estimate)	N/A	3/26	Google	Proprietary
II Rank II, #4 to 17	gemini-3.6-flash	1113	-24/+6	216	N / A	L - (estimate)	N/A	7/26	Google	Proprietary
II Rank II, #4 to 17	gpt-5.6-terra	1109	-11/+3	1690	N / A	L - (estimate)	N/A	7/26	OpenAI	Proprietary
II Rank II, #4 to 17	gpt-5.5	1103	-11/+3	2762	N / A	L - (estimate)	N/A	4/26	OpenAI	Proprietary
II Rank II, #4 to 17	gpt-5.6-luna	1102	-12/+5	1444	N / A	L - (estimate)	N/A	7/26	OpenAI	Proprietary
II Rank II, #4 to 17	gemini-3-7-flash	1100	-15/+5	991	N / A	L - (estimate)	N/A	8/26	Google	Proprietary
II Rank II, #4 to 17	gemma-4-31b-it	1099	-10/+5	3665	117 mWh	S - 31 billion	Dense	4/26	Google	Open-weight
II Rank II, #4 to 17	gemini-3-8-flash	1098	-13/+8	1182	N / A	L - (estimate)	N/A	9/26	Google	Proprietary
II Rank II, #4 to 17	gemini-3.5-flash-lite	1094	-19/+11	598	N / A	L - (estimate)	N/A	7/26	Google	Proprietary
II Rank II, #4 to 17	gemini-3.5-flash	1092	-14/+10	1098	N / A	XL - (estimate)	N/A	5/26	Google	Proprietary
III Rank III, #18 to 36	qwen3-max-2025-09-23	1090	-10/+8	2665	N / A	XL - (estimate)	N/A	9/25	Qwen	Proprietary
III Rank III, #18 to 36	gpt-5.4	1087	-12/+9	2069	N / A	L - (estimate)	N/A	3/26	OpenAI	Proprietary
III Rank III, #18 to 36	claude-haiku-5-5	1086	-23/+14	456	N / A	L - (estimate)	N/A	10/26	Anthropic	Proprietary
III Rank III, #18 to 36	gpt-5.1	1084	-17/+14	730	N / A	L - (estimate)	N/A	11/25	OpenAI	Proprietary
III Rank III, #18 to 36	gemini-3.1-pro-preview	1083	-10/+8	2960	N / A	XL - (estimate)	N/A	2/26	Google	Proprietary
III Rank III, #18 to 36	gpt-5.2	1080	-16/+11	1067	N / A	L - (estimate)	N/A	12/25	OpenAI	Proprietary
III Rank III, #18 to 36	gemini-2.5-flash	1079	-11/+7	3063	N / A	XL - (estimate)	N/A	6/25	Google	Proprietary
III Rank III, #18 to 36	gpt-5.4-mini	1077	-13/+8	2188	N / A	S - (estimate)	N/A	3/26	OpenAI	Proprietary
III Rank III, #18 to 36	deepseek-v3-0324	1074	-11/+8	3622	3979 mWh	XL - 685 billion	MoE	3/25	DeepSeek	Open-weight
III Rank III, #18 to 36	gemma-4-26b-a4b-it	1074	-10/+8	3409	84 mWh	S - 26 billion	MoE	4/26	Google	Open-weight
III Rank III, #18 to 36	minimax-m3	1072	-14/+13	1389	1718 mWh	XL - 428 billion	MoE	6/26	MiniMax	Open-weight
III Rank III, #18 to 36	kimi-k2	1071	-12/+12	1562	3785 mWh	XL - 1000 billion	MoE	9/25	Moonshot AI	Open-weight
III Rank III, #18 to 36	gpt-6-sol	1070	-18/+18	693	N / A	XL - (estimate)	N/A	9/26	OpenAI	Proprietary
III Rank III, #18 to 36	mistral-large-4-0	1069	-22/+22	491	4445 mWh	XL - 1050 billion	MoE	10/26	Mistral AI	Open-weight
III Rank III, #18 to 36	mimo-v2-6-pro	1068	-19/+22	618	4173 mWh	XL - 1020 billion	MoE	9/26	Xiaomi	Open-weight
III Rank III, #18 to 36	claude-4-5-sonnet	1065	-8/+8	4389	N / A	XL - (estimate)	N/A	9/25	Anthropic	Proprietary
III Rank III, #18 to 36	magistral-medium	1064	-14/+14	1195	N / A	L - (estimate)	N/A	6/25	Mistral AI	Proprietary
III Rank III, #18 to 36	mistral-medium-3-5	1064	-10/+15	1640	939 mWh	L - 128 billion	Dense	4/26	Mistral AI	Open-weight
III Rank III, #18 to 36	mimo-v2-6-flash	1061	-17/+21	579	1562 mWh	L - 309 billion	MoE	9/26	Xiaomi	Open-weight
IV Rank IV, #37 to 50	gemma-3-27b	1061	-5/+9	6242	112 mWh	S - 27 billion	Dense	3/25	Google	Open-weight
IV Rank IV, #37 to 50	mistral-medium-2508	1059	-6/+8	5600	N / A	L - (estimate)	N/A	8/25	Mistral AI	Proprietary
IV Rank IV, #37 to 50	claude-5-sonnet	1059	-11/+15	1219	N / A	XL - (estimate)	N/A	6/26	Anthropic	Proprietary
IV Rank IV, #37 to 50	claude-4-sonnet	1056	-8/+13	2238	N / A	XL - (estimate)	N/A	5/25	Anthropic	Proprietary
IV Rank IV, #37 to 50	gpt-4.1-mini	1054	-6/+8	4968	N / A	M - (estimate)	N/A	4/25	OpenAI	Proprietary
IV Rank IV, #37 to 50	gemini-3-pro-preview	1054	-16/+21	737	N / A	XL - (estimate)	N/A	11/25	Google	Proprietary
IV Rank IV, #37 to 50	deepseek-v4-pro	1051	-9/+10	2522	8890 mWh	XL - 1600 billion	MoE	4/26	DeepSeek	Open-weight
IV Rank IV, #37 to 50	qwen3-coder-next	1045	-19/+14	881	332 mWh	M - 80 billion	MoE	2/26	Qwen	Open-weight
IV Rank IV, #37 to 50	mistral-large-2512	1044	-9/+6	4803	4134 mWh	XL - 675 billion	MoE	12/25	Mistral AI	Open-weight
IV Rank IV, #37 to 50	deepseek-chat-v3.1	1043	-15/+14	1150	3979 mWh	XL - 685 billion	MoE	8/25	DeepSeek	Open-weight
IV Rank IV, #37 to 50	gpt-5-mini	1043	-13/+11	1699	N / A	S - (estimate)	N/A	8/25	OpenAI	Proprietary
IV Rank IV, #37 to 50	o4-mini	1040	-14/+10	1542	N / A	S - (estimate)	N/A	4/25	OpenAI	Proprietary
IV Rank IV, #37 to 50	gpt-5.4-nano	1039	-13/+11	2071	N / A	XS - (estimate)	N/A	3/26	OpenAI	Proprietary
IV Rank IV, #37 to 50	Qwen3-Coder-480B-A35B-Instruct	1038	-18/+18	623	1951 mWh	XL - 480 billion	MoE	7/25	Qwen	Open-weight
V Rank V, #51 to 66	mistral-small-2603	1037	-10/+8	2842	347 mWh	L - 119 billion	MoE	3/26	Mistral AI	Open-weight
V Rank V, #51 to 66	deepseek-v4-flash	1035	-11/+10	1778	1524 mWh	L - 284 billion	MoE	4/26	DeepSeek	Open-weight
V Rank V, #51 to 66	command-a	1032	-10/+7	4812	857 mWh	L - 111 billion	Dense	3/25	Cohere	Open-weight
V Rank V, #51 to 66	gemma-3-12b	1031	-8/+6	6075	94 mWh	XS - 12 billion	Dense	3/25	Google	Open-weight
V Rank V, #51 to 66	gpt-5	1029	-14/+12	1358	N / A	L - (estimate)	N/A	8/25	OpenAI	Proprietary
V Rank V, #51 to 66	DeepSeek-V3.2	1029	-10/+11	2057	3979 mWh	XL - 685 billion	MoE	12/25	DeepSeek	Open-weight
V Rank V, #51 to 66	trinity-large-preview	1027	-14/+15	1097	1524 mWh	L - 398 billion	MoE	1/26	Arcee	Open-weight
V Rank V, #51 to 66	mistral-small-2506	1027	-6/+9	3958	109 mWh	S - 24 billion	Dense	6/25	Mistral AI	Open-weight
V Rank V, #51 to 66	gpt-6-luna	1025	-18/+21	729	N / A	L - (estimate)	N/A	9/26	OpenAI	Proprietary
V Rank V, #51 to 66	kimi-k2.6	1021	-14/+17	1057	3785 mWh	XL - 1000 billion	MoE	4/26	Moonshot AI	Open-weight
V Rank V, #51 to 66	glm-5-3	1017	-16/+15	974	4095 mWh	XL - 744 billion	MoE	8/26	Zhipu	Open-weight
V Rank V, #51 to 66	inkling	1017	-16/+15	941	4134 mWh	XL - 975 billion	MoE	7/26	Thinking Machines	Open-weight
V Rank V, #51 to 66	glm-4.5	1017	-18/+24	588	1892 mWh	L - 355 billion	MoE	7/25	Zhipu	Open-weight
V Rank V, #51 to 66	llama-maverick	1016	-6/+8	4058	1601 mWh	XL - 400 billion	MoE	4/25	Meta	Open-weight
V Rank V, #51 to 66	trinity-large-thinking	1014	-12/+14	1470	1524 mWh	L - 398 billion	MoE	4/26	Arcee	Open-weight
V Rank V, #51 to 66	kimi-k3	1014	-16/+19	1009	26322 mWh	XL - 2800 billion	MoE	7/26	Moonshot AI	Open-weight
VI Rank VI, #67 to 82	llama-4-scout	1003	-10/+4	6353	400 mWh	L - 109 billion	MoE	4/25	Meta	Open-weight
VI Rank VI, #67 to 82	magistral-small-2506	1003	-12/+7	2319	109 mWh	S - 24 billion	Dense	6/25	Mistral AI	Open-weight
VI Rank VI, #67 to 82	glm-5.2	1002	-18/+13	875	4095 mWh	XL - 744 billion	MoE	6/26	Zhipu	Open-weight
VI Rank VI, #67 to 82	kimi-k2-thinking	1001	-22/+15	749	3785 mWh	XL - 1000 billion	MoE	11/25	Moonshot AI	Open-weight
VI Rank VI, #67 to 82	gemma-3-4b	1001	-6/+7	6892	84 mWh	XS - 4 billion	Dense	3/25	Google	Open-weight
VI Rank VI, #67 to 82	glm-5-3-flash	1000	-15/+18	874	1621 mWh	L - 320 billion	MoE	8/26	Zhipu	Open-weight
VI Rank VI, #67 to 82	gemma-3n-e4b-it	996	-9/+8	4084	84 mWh	XS - 8 billion	Matformer	5/25	Google	Open-weight
VI Rank VI, #67 to 82	glm-4.6	996	-25/+23	484	1892 mWh	L - 357 billion	MoE	9/25	Zhipu	Open-weight
VI Rank VI, #67 to 82	kimi-k2.5	994	-13/+11	1838	3785 mWh	XL - 1000 billion	MoE	1/26	Moonshot AI	Open-weight
VI Rank VI, #67 to 82	lfm2-8b-a1b	993	-15/+13	1493	81 mWh	XS - 8 billion	MoE	10/25	Liquid	Open-weight
VI Rank VI, #67 to 82	gpt-4.1-nano	992	-9/+9	3959	N / A	S - (estimate)	N/A	4/25	OpenAI	Proprietary
VI Rank VI, #67 to 82	deepseek-r1	991	-13/+12	2315	3979 mWh	XL - 671 billion	MoE	1/25	DeepSeek	Open-weight
VI Rank VI, #67 to 82	llama-3.1-nemotron-70b-instruct	987	-11/+7	4920	658 mWh	M - 70 billion	Dense	10/24	Nvidia	Open-weight
VI Rank VI, #67 to 82	mistral-saba	983	-13/+7	3257	N / A	S - (estimate)	N/A	2/25	Mistral AI	Proprietary
VI Rank VI, #67 to 82	apertus-v1-5-70b	983	-24/+18	623	658 mWh	M - 70 billion	Dense	7/27	Swiss AI	Open source
VI Rank VI, #67 to 82	qwen3-8-flash	982	-22/+18	723	347 mWh	L - 125 billion	MoE	8/26	Qwen	Open-weight
VII Rank VII, #83 to 96	llama-3.3-70b	980	-9/+7	7155	658 mWh	M - 70 billion	Dense	12/24	Meta	Open-weight
VII Rank VII, #83 to 96	hermes-4-70b	979	-13/+10	2549	658 mWh	M - 70 billion	Dense	8/25	Nous	Open-weight
VII Rank VII, #83 to 96	deepseek-r1-0528	978	-15/+14	1431	3979 mWh	XL - 685 billion	MoE	5/25	DeepSeek	Open-weight
VII Rank VII, #83 to 96	aya-expanse-32b	978	-9/+11	3479	118 mWh	S - 32 billion	Dense	12/24	Cohere	Open-weight
VII Rank VII, #83 to 96	muse-glimmer-30b	977	-14/+17	1489	116 mWh	S - 30 billion	Dense	8/26	Meta	Open-weight
VII Rank VII, #83 to 96	Qwen3-8-2-4T-A95B	975	-15/+20	992	24924 mWh	XL - 2400 billion	MoE	8/26	Qwen	Open-weight
VII Rank VII, #83 to 96	qwen3.7-max	972	-18/+21	692	N / A	XL - (estimate)	N/A	5/26	Qwen	Proprietary
VII Rank VII, #83 to 96	glm-5.1	967	-14/+17	1026	4095 mWh	XL - 744 billion	MoE	4/26	Zhipu	Open-weight
VII Rank VII, #83 to 96	qwen3-32b	967	-20/+26	455	118 mWh	S - 32 billion	Dense	4/25	Qwen	Open-weight
VII Rank VII, #83 to 96	EuroLLM-22B-Instruct-2512	967	-11/+17	1717	106 mWh	S - 22 billion	Dense	12/25	EuroLLM	Open source
VII Rank VII, #83 to 96	lfm2-24b-a2b	963	-12/+17	1451	82 mWh	S - 24 billion	MoE	2/26	Liquid	Open-weight
VII Rank VII, #83 to 96	qwen-3-8-27b	960	-13/+15	1343	112 mWh	S - 27 billion	Dense	8/26	Qwen	Open-weight
VII Rank VII, #83 to 96	Apertus-70B-Instruct-2509	960	-9/+11	2917	658 mWh	M - 70 billion	Dense	9/25	Swiss AI	Open source
VII Rank VII, #83 to 96	inkling-small	958	-14/+18	915	1504 mWh	L - 276 billion	Dense	7/26	Thinking Machines	Open-weight
VIII Rank VIII, #97 to 121	llama-3.1-405b	956	-6/+8	7937	9134 mWh	XL - 405 billion	Dense	7/24	Meta	Open-weight
VIII Rank VIII, #97 to 121	gpt-oss-120b	956	-6/+10	3894	342 mWh	L - 117 billion	MoE	8/25	OpenAI	Open-weight
VIII Rank VIII, #97 to 121	glm-5	953	-10/+15	1146	4095 mWh	XL - 744 billion	MoE	2/26	Zhipu	Open-weight
VIII Rank VIII, #97 to 121	deepseek-v4-1-flash	950	-15/+21	779	3164 mWh	XL - 552 billion	MoE	9/26	DeepSeek	Open-weight
VIII Rank VIII, #97 to 121	qwen3.6-plus	950	-10/+16	1230	N / A	XL - (estimate)	N/A	4/26	Qwen	Proprietary
VIII Rank VIII, #97 to 121	minimax-m2	947	-12/+19	694	733 mWh	L - 230 billion	MoE	10/25	MiniMax	Open-weight
VIII Rank VIII, #97 to 121	command-a-plus	943	-16/+25	355	878 mWh	L - 218 billion	MoE	9/26	Cohere	Open-weight
VIII Rank VIII, #97 to 121	gpt-5-nano	942	-11/+13	1226	N / A	XS - (estimate)	N/A	4/25	OpenAI	Proprietary
VIII Rank VIII, #97 to 121	minimax-m2.5	941	-10/+14	1178	733 mWh	L - 229 billion	MoE	2/26	MiniMax	Open-weight
VIII Rank VIII, #97 to 121	qwen3.5-397b-a17b	940	-7/+9	1802	1601 mWh	L - 397 billion	MoE	2/26	Qwen	Open-weight
VIII Rank VIII, #97 to 121	minimax-m2.7	935	-10/+12	1076	733 mWh	L - 230 billion	MoE	3/26	MiniMax	Open-weight
VIII Rank VIII, #97 to 121	apertus-v1-5-8b	933	-13/+18	620	89 mWh	XS - 8 billion	Dense	7/26	Swiss AI	Open source
VIII Rank VIII, #97 to 121	granite-4.1-8b	932	-9/+13	1111	89 mWh	XS - 8 billion	Dense	4/26	IBM	Open-weight
VIII Rank VIII, #97 to 121	glm-4.7	926	-8/+12	957	1892 mWh	L - 357 billion	MoE	12/25	Zhipu	Open-weight
VIII Rank VIII, #97 to 121	phi-4	926	-2/+8	6664	96 mWh	XS - 14 billion	Dense	12/24	Microsoft	Open-weight
VIII Rank VIII, #97 to 121	nemotron-3-super-120b-a12b	924	-6/+12	1302	376 mWh	L - 120 billion	MoE	3/26	Nvidia	Open source
VIII Rank VIII, #97 to 121	deepseek-r1-distill-llama-70b	910	-6/+6	1693	658 mWh	M - 70 billion	Dense	1/25	DeepSeek	Open-weight
VIII Rank VIII, #97 to 121	llama-3.1-8b	907	-4/+3	7502	89 mWh	XS - 8 billion	Dense	7/24	Meta	Open-weight
VIII Rank VIII, #97 to 121	gpt-oss-20b	907	-4/+4	5321	83 mWh	S - 21 billion	MoE	8/25	OpenAI	Open-weight
VIII Rank VIII, #97 to 121	nemotron-3-5-lightning	906	-6/+9	1190	83 mWh	S - 30 billion	MoE	8/26	Nvidia	Open source
VIII Rank VIII, #97 to 121	qwen3-30b-a3b	901	-4/+10	1058	83 mWh	S - 30 billion	MoE	5/25	Qwen	Open-weight
VIII Rank VIII, #97 to 121	qwen3.5-35b-a3b	887	-3/+4	2136	166 mWh	S - 35 billion	MoE	2/26	Qwen	Open-weight
VIII Rank VIII, #97 to 121	qwen-3-8b	853	-4/+6	205	89 mWh	XS - 8 billion	Dense	7/25	Qwen	Open-weight
VIII Rank VIII, #97 to 121	qwq-32b	848	-2/+3	985	118 mWh	S - 32 billion	Dense	4/25	Qwen	Open-weight
VIII Rank VIII, #97 to 121	olmo-3-32b-think	714	-0/+0	593	118 mWh	S - 32 billion	Dense	11/25	Ai2	Open source