Machine Translation (other-language-to-English) on Neko (test) benchmark leaderboard · SOTA2 Research

Machine Translation (other-language-to-English) on Neko (test)

22.34EXACTNeoAMT-8B

Evaluation Results

Method	Links
NeoAMT-8B2026.01		22.34	26.24	25.84	28.67	78.28	66.4
Qwen3-32B2026.01		19.65	22.21	22.75	24.36	76.68	63.72
NeoAMT-4B + process reward2026.01		19.25	23.55	23.55	27.19	74.06	64.43
GPT-OSS-20B2026.01		18.84	22.07	21.67	23.42	73.96	62.62
NeoAMT-8B + process reward2026.01		18.57	21.13	23.82	26.78	79.5	67.58
Qwen3-14B2026.01		18.3	20.05	21.53	22.34	76.91	63.56
NeoAMT-4B2026.01		17.63	19.11	20.05	21.53	72.93	58.16
GRPO-8B2026.01		17.63	21.13	20.19	22.75	72.84	61.11
Qwen3-8B2026.01		17.36	19.11	20.32	21.13	71.24	58.13
MT-R1-Zero-7B-Mix2026.01		16.29	18.3	18.84	19.11	67.22	53.8
Hunyuan-MT-7B2026.01		16.15	17.63	19.78	20.19	69.46	55.81
SFT-8B2026.01		15.61	18.17	19.11	20.32	72.39	60.01
Aya-expanse-8B2026.01		14.54	17.22	17.63	19.38	70.2	56.18
SFT-4B2026.01		13.73	16.96	17.23	18.84	66.92	54
GRPO-4B2026.01		13.73	15.74	16.82	18.98	71.29	55.34
X-ALMA2026.01		13.46	15.07	16.55	17.09	71.28	58.87
mExTrans-7B2026.01		13.46	18.71	16.42	19.78	70.37	57.42
Qwen3-4B2026.01		13.19	16.02	17.23	18.57	65.24	51.94
Aya-23-8B2026.01		12.92	15.88	15.48	17.36	63.77	50.26
TowerInstruct-7B-v0.22026.01		11.71	14.4	14.27	16.15	60.2	47.49
Llama-3.1-8B-Instruct2026.01		11.44	13.46	13.19	14.54	60.64	47.79
ALMA-R-7B2026.01		10.09	13.86	13.06	15.75	62.56	48.78