HATEBENCH
● Models vs Hateful Hindi Text
ACCURACY DISTRIBUTION
SUCCESS RATE ON HINDI HATE SPEECH DEFINITIONS
01
gemini-3.1-flash-lite
75%
02
trinity-large-preview
71%
03
claude-opus-4.8
71%
04
gpt-5.2-pro
71%
05
hy3-preview
71%
06
gemini-3-flash-preview
70%
07
gemini-3.1-flash-lite-preview
70%
08
gemma-4-31b-it
70%
09
gemma-3n-e4b-it
70%
10
claude-fable-5
68%
11
claude-opus-4.7
67%
12
claude-opus-4.6
67%
13
claude-sonnet-4.6
67%
14
gpt-5.2
67%
15
gpt-5.5-pro
67%
16
deepseek-v4-flash
67%
17
gemini-3-pro-preview
67%
18
gemma-3-12b-it
67%
19
gpt-5.4-pro
67%
20
qwen3.5-plus-02-15
67%
21
gemini-3.5-flash
66%
22
gpt-5.3-chat
66%
23
kimi-k2.5
66%
24
grok-4.1-fast
66%
25
gemma-3-27b-it
65%
26
qwen3.7-max
65%
27
minimax-m3
65%
28
glm-5-turbo
65%
29
grok-4.3
65%
30
nova-2-lite-v1
65%
31
qwen3.6-max-preview
65%
32
claude-sonnet-4
64%
33
gpt-5.4
64%
34
gpt-5.4-nano
64%
35
command-a
63%
36
qwen3.5-plus-20260420
63%
37
qwen3.6-flash
63%
38
glm-5.1
63%
39
glm-4.7
63%
40
seed-2.0-lite
63%
41
mimo-v2-pro
63%
42
gpt-5.4-mini
62%
43
gpt-5.5
62%
44
sarvam-105b
61%
45
deepseek-v4-pro
61%
46
glm-5
61%
47
step-3.5-flash
61%
48
nemotron-3-ultra-550b-a55b
61%
49
mimo-v2-omni
60%
50
deepseek-chat-v3.1
59%
51
minimax-m2.1
59%
52
sarvam-30b
59%
53
qwen3-32b
59%
54
ministral-14b-2512
59%
55
deepseek-v3.2
58%
56
llama-3.3-nemotron-super-49b-v1.5
58%
57
qwen3-235b-a22b-2507
57%
58
mistral-medium-3-5
56%
59
mistral-small-2603
55%
60
qwen3.5-9b
54%
61
minimax-m2.7
54%
62
minimax-m2.5
53%
63
llama-4-maverick
51%
64
hunyuan-a13b-instruct
51%
65
mercury-2
49%
66
phi-4
47%
67
molmo-2-8b
43%