HATEBENCH
● Models vs Hateful Hindi Text
ACCURACY DISTRIBUTION
SUCCESS RATE ON HINDI HATE SPEECH DEFINITIONS
01
trinity-large-preview
71%
02
gpt-5.2-pro
71%
03
gemini-3-flash-preview
70%
04
gemini-3.1-flash-lite-preview
70%
05
gemma-3n-e4b-it
70%
06
claude-opus-4.6
67%
07
claude-sonnet-4.6
67%
08
gpt-5.2
67%
09
gemini-3-pro-preview
67%
10
gemma-3-12b-it
67%
11
gpt-5.4-pro
67%
12
qwen3.5-plus-02-15
67%
13
gpt-5.3-chat
66%
14
kimi-k2.5
66%
15
grok-4.1-fast
66%
16
gemma-3-27b-it
65%
17
glm-5-turbo
65%
18
nova-2-lite-v1
65%
19
claude-sonnet-4
64%
20
gpt-5.4
64%
21
gpt-5.4-nano
64%
22
command-a
63%
23
glm-4.7
63%
24
seed-2.0-lite
63%
25
mimo-v2-pro
63%
26
gpt-5.4-mini
62%
27
sarvam-105b
61%
28
glm-5
61%
29
step-3.5-flash
61%
30
mimo-v2-omni
60%
31
deepseek-chat-v3.1
59%
32
minimax-m2.1
59%
33
sarvam-30b
59%
34
qwen3-32b
59%
35
ministral-14b-2512
59%
36
deepseek-v3.2
58%
37
llama-3.3-nemotron-super-49b-v1.5
58%
38
qwen3-235b-a22b-2507
57%
39
mistral-small-2603
55%
40
qwen3.5-9b
54%
41
minimax-m2.7
54%
42
minimax-m2.5
53%
43
llama-4-maverick
51%
44
hunyuan-a13b-instruct
51%
45
mercury-2
49%
46
phi-4
47%
47
molmo-2-8b
43%