Speech-to-Text on VoiceBench benchmark leaderboard · SOTA2 Research

Speech-to-Text on VoiceBench

4.78AlpacaEval ScoreGPT-4o-Audio

Evaluation Results

Method	Links
GPT-4o-Audio2026.02		4.78	4.49	4.58	75.5	80.25	89.23	84.1	76.02	98.65	86.75
Qwen2.5-Omni2026.02		4.49	3.93	—	55.71	61.32	81.1	—	52.87	99.42	—
Kimi-Audio2026.02		4.46	3.97	4.2	63.12	62.17	83.52	69.7	61.1	100	76.91
VITA-1.52026.02		4.21	3.66	3.48	38.88	52.15	71.65	55.3	38.14	97.69	64.53
Ultravox-v0.5-Llama-3.2-1b2025.11		4.02	3.53	3.48	46.38	29.99	33.41	51.4	43.51	97.88	—
LFM2-Audio-1.5B2025.11		3.78	3.48	3.12	34.81	33.99	45.49	51.2	30.13	98.85	—
Qwen2.5-Omni-3B2025.11		3.72	3.51	3.42	44.94	55.29	76.26	61.3	32.9	88.46	—
LLaMA-Omni2026.02		3.7	3.46	2.92	39.69	25.93	27.47	49.2	14.87	11.35	41.12
VITA-1.02026.02		3.38	2.15	1.87	27.94	25.7	29.01	47.7	22.82	26.73	36.43
Ex-Omni2026.02		2.57	2.87	2.19	40.14	24.24	25.49	50.4	16.22	83.08	43.57
Mini-Omni22026.02		2.32	2.18	1.79	9.31	24.27	26.59	46.4	11.56	57.5	33.49
Mini-Omni22025.11		2.32	2.18	1.79	9.31	24.27	26.59	46.4	11.56	57.5	—
Moshi2026.02		2.01	1.6	1.3	15.64	24.04	25.93	47.4	10.12	44.23	29.51
Moshi2025.11		2.01	1.6	1.3	15.64	24.04	25.93	47.4	10.12	44.23	—
Mini-Omni2026.02		1.95	2.02	1.61	13.92	24.69	26.59	46.3	13.58	37.12	30.42