Subtitle Translation on Subtitle Dataset EN-ZH benchmark leaderboard · SOTA2 Research

Subtitle Translation on Subtitle Dataset EN-ZH

92.4Translation ScoreGPT-5

Evaluation Results

Method	Links
GPT-52026.01		92.4	91.3	91.2	87	71.1
Qwen-Max2026.01		91.9	86.7	88.4	84.4	61.3
DeepSeek-V3.12026.01		91.2	84.3	92.3	85.3	63.5
Qwen2.5-14B (pi_st)2026.01		90.6	88.2	94.6	84.3	76.6
DeepSeek-R12026.01		90.5	86.4	90.3	85.7	70.8
Qwen-Max2026.01		90	85.3	91.3	85	66.8
DeepSeek-V3.12026.01		89.5	86.4	90.2	84.1	63
GPT-4o2026.01		89.3	82.7	87.3	82.3	59.8
GPT-52026.01		89.1	90.3	92.3	86.1	75.2
GPT-4o2026.01		88.5	87.1	90.4	83	64.6
DeepSeek-R12026.01		88.5	87.2	93.4	85.6	73.5
Qwen2.5-14B (pi_st)2026.01		88.3	88.4	96.4	86.8	81.7
Qwen2.5-14B (pi_sft)2026.01		86.4	89.7	95.3	82	59.1
Qwen2.5-14B (pi_sft)2026.01		85.2	91.2	95.3	80.1	54.8
Google Translate2026.01		84.2	—	—	79.7	54.4
Gold Reference2026.01		83.6	—	—	82.6	71.5
Gold Reference2026.01		83	—	—	80.3	73.3
Google Translate2026.01		79.8	—	—	66.3	50.2
NLLB-3.3B2026.01		61.4	—	—	54	43.7
MADLAD-10B2026.01		59.7	—	—	55.5	46.3
VideoDubber2026.01		53.6	—	—	54.8	50.1
VideoDubber2026.01		46.9	—	—	51.9	49.7
MADLAD-10B2026.01		45.1	—	—	38.9	37.6
NLLB-3.3B2026.01		29.1	—	—	21.7	20.8