Text Embedding on AfriMTEB Lite (test) benchmark leaderboard · SOTA2 Research

Text Embedding on AfriMTEB Lite (test)

55Hate Speech Scoregemini embedding 001

Evaluation Results

Method	Links
gemini embedding 0012025.10		55	53.8	75.3	83.6	35.5	88.1	84.2	83.9	76.8	17.7	69.2	34.6	63.1
AfriE5-large-instruct2025.10		51.7	50.7	69	77.7	32.8	91.2	92	75.4	79.5	26.2	72	45.7	63.7
mE5-large-instruct2025.10		51.5	47	64.5	75.7	31.5	91.4	91.5	75.5	78.8	22	71.2	43.9	62
GritLM-7B2025.10		51.4	45.6	59.8	55	29.6	46.5	52	70.8	75.3	8.8	50.8	22.2	47.3
Linq-Embed-Mistral2025.10		51	43.8	59.7	52.1	30	46.7	52	70.3	76.5	7.6	50.7	22.1	46.9
Qwen3-Embedding 8B2025.10		50.8	46.2	58.8	68.3	29.1	58.3	57.3	67.7	77.6	8	53.5	20.9	49.7
bge-m32025.10		50.1	47.9	68.7	69.9	29.4	78.1	80.4	75.4	72.7	10.2	55.6	21	55
Qwen3-Embedding-4B2025.10		50	41.2	56.8	58.3	28.3	47.7	49	70.1	76	7.9	49.2	17.5	46
mE5-large2025.10		49.8	48.9	65.2	69.9	31.3	86.9	88.7	77.1	77.7	11.6	60.4	25.6	57.8
SFR-Embedding-Mistral2025.10		49.3	44.5	58.9	50.9	28.8	47.7	53	62.5	77	8.5	49.7	—	46.1
E5 mistral 7b instruct2025.10		49	45.7	58.7	50.2	28.6	47.9	53.3	61.8	76	7.2	49.2	21.8	45.8
Qwen3-Embedding 0.6B2025.10		48.6	41.5	58.1	46.3	28.1	33	38.1	68.5	70.5	3.5	40	12	40.7
KaLM2025.10		48.5	44.3	61	51	28.9	53.2	58.6	63.4	74.6	7.3	48.9	16.8	46.4
mmBERT-base2025.10		48.2	39.9	54	4	27.3	1.8	2.2	72.5	55.1	2.4	34.2	5	28.9
gte-Qwen2-7B-instruct2025.10		46	43.5	58.6	55.6	27.8	58.4	60	57.1	79.8	10.4	50.8	22.4	47.5
EmbeddingGemma 300m2025.10		45.4	39	53.6	10.6	26.6	12.2	17.9	49.4	59	1.4	29.6	4.6	29.1