Language Model Evaluation across Subsets on ELEPHANT benchmark leaderboard · SOTA2 Research

Language Model Evaluation across Subsets on ELEPHANT

93OEQ Validation ScoreQwen3-80B

Evaluation Results

Method	Links
Qwen3-80B2026.01		93	55	8	85	28	18	20
Gemini-2.5-Pro2026.01		89	55	21	77	36	33	29
Qwen3-80B2026.01		71	31	1	82	20	14	-6
Qwen3-80B2026.01		68	30	-9	75	13	-35	-42
Gemini-2.5-Pro2026.01		67	33	-22	76	36	-27	-23
GPT-4o2026.01		67	72	-9	58	75	-3	-16
Claude-Sonnet-42026.01		60	58	24	71	33	29	25
Llama-8B2026.01		59	73	30	58	75	35	32
GPT-4o2026.01		57	67	34	70	66	41	34
Llama-70B2026.01		56	73	30	51	44	40	31
GPT-4o2026.01		56	78	34	76	87	34	34
Gemini-2.5-Pro2026.01		49	36	10	44	37	20	33
Claude-Sonnet-42026.01		46	41	-20	49	19	-21	-19
Claude-Sonnet-42026.01		44	42	6	56	26	9	-6
Llama-8B2026.01		42	69	31	38	56	42	24
Llama-70B2026.01		41	72	-33	23	84	-20	-40
Llama-70B2026.01		40	66	31	41	52	45	30
Llama-8B2026.01		33	66	5	16	64	-1	-27