Speech-to-Text reasoning on Kimi-Audio-Evalkit s2t benchmark leaderboard · SOTA2 Research

Speech-to-Text reasoning on Kimi-Audio-Evalkit s2t

87Llama Q ScoreGemini 2.5-Pro

Evaluation Results

Method	Links
Gemini 2.5-Pro2026.06		87	79	93	3.73	3.08	4.49	90	4.69	83.6
Qwen3-Omni-30B-A3B2026.06		82	64	70	3.95	3.65	3.86	68	4.98	76.5
Fun-Audio-Chat-8B2026.06		80	61	63	4.2	3.67	3.81	59	4.99	74.5
FlexiSLM-7B-Stage32026.06		80	61	62	3.84	3.63	3.22	62	5	72.4
FlexiSLM-7B-Stage32026.06		80	61	62	3.82	3.59	3.22	63	5	72.3
Mimo-Audio-7B2026.06		79	61	57	3.83	3.68	3.24	54	4.97	70.6
Kimi-Audio-7B2026.06		79	66	62	3.38	3	3.26	58	4.98	69.7
FlexiSLM-7B-Stage22026.06		79	61	61	3.74	3.39	3.18	58	4.99	70.6
FlexiSLM-7B-Stage22026.06		79	61	61	3.74	3.39	3.18	58	4.99	70.6
FlexiSLM-7B-Stage32026.06		79	58	60	3.7	3.49	3.17	61	4.97	70.6
FlexiSLM-7B-Stage22026.06		78	58	59	3.67	3.34	3.1	57	4.98	69.2
FlexiSLM-7B-Stage22026.06		78	58	59	3.67	3.34	3.1	57	4.98	69.2
FlexiSLM-7B-Stage32026.06		78	58	60	3.7	3.43	3.11	61	4.96	70.2
FlexiSLM-7B-Stage32026.06		77	55	58	3.7	3.35	3.14	59	4.97	69
GLM-4-Voice-9B2026.06		76	50	51	3.24	2.95	2.32	48	4.74	61.1
Qwen2.5-Omni-7B2026.06		76	60	58	3.3	3.35	2.68	54	4.92	66.7
FlexiSLM-7B-Stage32026.06		75	53	56	3.6	3.28	2.94	58	4.97	67.2
Qwen2-Audio-7B2026.06		68	44	39	2.87	2.82	2.25	31	4.89	54.9