Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Safety Evaluation on XSTest
Loading...
98.4
Safety Score
Base
32.464
49.582
66.7
83.818
Jan 22, 2025
Apr 1, 2025
Jun 10, 2025
Aug 19, 2025
Oct 28, 2025
Jan 6, 2026
Mar 17, 2026
Safety Score
Updated 1mo ago
Evaluation Results
Method
Method
Links
Safety Score
Base
2025.09
98.4
Qwen2.5 Instruct (72B)
Evaluation Source=HELM
2025.01
97.9
GPT-4o (2024-05-13)
Evaluation Source=HELM
2025.01
97.3
DeepSeek-V3
Evaluation Source=HELM...
2025.01
97.1
o1 (2024-12-17)
Evaluation Source=HELM
2025.01
97
Claude-3.7-Sonnet
Evaluation Source=HELM
2025.01
96.4
DeepSeek-R1
Evaluation Source=HELM...
2025.01
95.3
DeepSeek-R1
Evaluation Source=HELM...
2025.01
94.4
IQuest-Coder-V1-40B-Thinking
Parameters=40B, Type=T...
2026.03
94.3
Qwen2.5-Coder-32B-Instruct
Parameters=32B, Type=I...
2026.03
90.6
Qwen3-Coder-480B-A35B-Instruct
Parameters=480B-A35B,...
2026.03
90.1
IQuest-Coder-V1-40B-Instruct
Parameters=40B, Type=I...
2026.03
89.3
Self-Improving Pretraining
Pre-training Data=RedP...
2026.01
88.4
TARS
2025.09
88.3
Llama Pretrain Baseline
Pre-training Data=RedP...
2026.01
87.6
GRPO
2025.09
86.8
Llama Base
Pre-training Strategy=...
2026.01
85.2
BackTrack
2025.09
80
IPO
2025.09
80
STAR
2025.09
76.9
Self-Improving Pretraining
Pre-training Data=RedP...
2026.01
49
Llama Base
Pre-training Strategy=...
2026.01
39.5
Llama Pretrain Baseline
Pre-training Data=RedP...
2026.01
35
Feedback
Search any
task
Search any
task