Dashboard
Adversarial prompt-injection benchmark results
4,000tests·308leaks (7.7%)·2,899defended·793errors·10runs
Leaderboard
View full →| # | Model | Elo↓ | Atk Rate | Def Rate |
|---|---|---|---|---|
| 1 | DeepSeekV32 | 1540 | 20.5% | 84.5% |
| 2 | GPT54 | 1529 | 9.0% | 69.5% |
| 3 | Gem3Flash | 1526 | 22.0% | 71.5% |
| 4 | KimiK25 | 1518 | 10.5% | 85.0% |
| 5 | ClaudeSonnet46 | 1507 | 7.5% | 64.5% |
| 6 | GemPro31 | 1502 | 6.0% | 65.0% |
| 7 | GPT5Nano | 1500 | 5.0% | 81.0% |
| 8 | GPT53Codex | 1498 | 2.5% | 65.5% |
| 9 | MiniMaxM25 | 1497 | 4.0% | 85.0% |
| 10 | GrokFast | 1491 | 4.5% | 87.0% |
| 11 | GPToss120B | 1488 | 0.5% | 88.5% |
| 12 | Grok4 | 1487 | 1.0% | 68.0% |
| 13 | GLM5 | 1486 | 5.5% | 86.5% |
| 14 | ClaudeOpus | 1486 | 5.5% | 68.0% |
| 15 | Gem25FlashLite | 1485 | 13.0% | 79.0% |
| 16 | TrinityLarge | 1476 | 10.5% | 63.0% |
| 17 | ClaudeSonnet45 | 1476 | 5.0% | 52.5% |
| 18 | ClaudeHaiku | 1453 | 1.0% | 76.0% |
| 19 | Gem20Flash | 1445 | 10.0% | 53.5% |
| 20 | Gem25Flash | 1436 | 10.5% | 56.0% |
Attack Success Heatmapattacker (row) vs defender (col)
ClaudeHaiku
ClaudeOpus
ClaudeSonnet45
ClaudeSonnet46
DeepSeekV32
GLM5
GPT53Codex
GPT54
GPT5Nano
GPToss120B
Gem20Flash
Gem25Flash
Gem25FlashLite
Gem3Flash
GemPro31
Grok4
GrokFast
KimiK25
MiniMaxM25
TrinityLarge
Atk Rate
ClaudeHaiku
0/100%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
0/100%
1%
ClaudeOpus
0/100%
0/100%
1/1010%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
3/1030%
1/1010%
1/1010%
3/1030%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
6%
ClaudeSonnet45
1/1010%
0/100%
1/1010%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
5/1050%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
5%
ClaudeSonnet46
1/1010%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
5/1050%
1/1010%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
4/1040%
8%
DeepSeekV32
0/100%
0/100%
2/1020%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
7/1070%
9/1090%
5/1050%
7/1070%
0/100%
0/100%
0/100%
0/100%
1/1010%
10/10100%
21%
GLM5
1/1010%
0/100%
1/1010%
2/1020%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
1/1010%
0/100%
1/1010%
0/100%
1/1010%
0/100%
3/1030%
0/100%
0/100%
6%
GPT53Codex
1/1010%
1/1010%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
3%
GPT54
0/100%
1/1010%
2/1020%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
3/1030%
3/1030%
1/1010%
2/1020%
0/100%
0/100%
1/1010%
0/100%
0/100%
4/1040%
9%
GPT5Nano
0/100%
0/100%
3/1030%
2/1020%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
3/1030%
0/100%
0/100%
1/1010%
0/100%
0/100%
0/100%
5%
GPToss120B
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
1%
Gem20Flash
0/100%
0/100%
4/1040%
0/100%
0/100%
2/1020%
0/100%
0/100%
0/100%
0/100%
3/1030%
4/1040%
1/1010%
0/100%
0/100%
0/100%
1/1010%
0/100%
1/1010%
4/1040%
10%
Gem25Flash
1/1010%
0/100%
2/1020%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
8/1080%
2/1020%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
4/1040%
11%
Gem25FlashLite
1/1010%
0/100%
3/1030%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
7/1070%
6/1060%
1/1010%
3/1030%
0/100%
0/100%
0/100%
0/100%
1/1010%
4/1040%
13%
Gem3Flash
1/1010%
0/100%
2/1020%
1/1010%
4/1040%
0/100%
0/100%
0/100%
0/100%
0/100%
7/1070%
7/1070%
8/1080%
2/1020%
0/100%
0/100%
0/100%
1/1010%
1/1010%
10/10100%
22%
GemPro31
0/100%
0/100%
2/1020%
2/1020%
0/100%
0/100%
0/100%
1/1010%
0/100%
0/100%
3/1030%
1/1010%
0/100%
2/1020%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
6%
Grok4
0/100%
0/100%
1/1010%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
1%
GrokFast
0/100%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
3/1030%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
0/100%
2/1020%
5%
KimiK25
0/100%
1/1010%
2/1020%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
6/1060%
0/100%
5/1050%
0/100%
0/100%
0/100%
0/100%
0/100%
5/1050%
11%
MiniMaxM25
0/100%
0/100%
2/1020%
2/1020%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
0/100%
1/1010%
0/100%
0/100%
0/100%
0/100%
0/100%
1/1010%
4%
TrinityLarge
1/1010%
1/1010%
3/1030%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
0/100%
3/1030%
7/1070%
0/100%
4/1040%
0/100%
0/100%
0/100%
0/100%
0/100%
2/1020%
11%
Def Rate
96%
98%
83%
95%
96%
99%
100%
100%
100%
100%
74%
68%
91%
83%
100%
99%
99%
97%
97%
75%
Best Attackers
1Gem3Flash
44/200 leaks landed
2DeepSeekV32
41/200 leaks landed
3Gem25FlashLite
26/200 leaks landed
4KimiK25
21/200 leaks landed
5TrinityLarge
21/200 leaks landed
Best Defenders
1GPToss120B
177/200 defended
2GrokFast
174/200 defended
3GLM5
173/200 defended
4KimiK25
170/200 defended
5MiniMaxM25
170/200 defended