LLM Security Rankings

Which LLM is the most vulnerable? View the stats from the largest researcher-validated exploit feed on the internet. 0DIN continuously red-teams production LLMs and publishes the results publicly. Every score reflects real attack attempts, not self-reported safety claims.

Live data • Last updated about 18 hours ago • 36 scans across 29 models

The five most jailbroken model families over the last six months, ranked by susceptibility to known vulnerability probes.

1 Grok
2 Flash
3 Haiku
4 GPT
5 Sonnet

How we rank

Get monthly LLM
rankings in your inbox

Comprehensive security rankings

Every frontier and open model, scored against researcher-validated exploits. Sort by risk, filter by vendor, and find the models your security team needs to watch most.

Live data • Last updated about 18 hours ago • 36 scans across 29 models

Model

Vendor

Sort by

Show retired models

#	Model	Risk	ASR%	Successful attacks
1	GPT-5.4	High	36%	492 of 1,385 probes
2	Grok 4.20	High	34%	465 of 1,385 probes
3	Gemini 3.5 Flash	Moderate	18%	309 of 1,683 probes
4	GPT-5 mini	Moderate	17%	278 of 1,598 probes
5	Claude 4.5 Haiku	Moderate	14%	187 of 1,385 probes
6	Claude 4.6 Sonnet	Moderate	11%	185 of 1,683 probes
7	Grok 4.5	Moderate	10%	1 of 10 probes
8	Claude 4.6 Opus	Low	9%	143 of 1,683 probes
9	Claude 4.7 Opus	Low	8%	130 of 1,683 probes
10	Claude 5 Sonnet	Low	7%	125 of 1,683 probes
11	Claude 4.8 Opus	Low	5%	87 of 1,697 probes
12	Claude 4.5 Sonnet	Low	5%	85 of 1,683 probes
13	GPT-5 Pro	Low	4%	58 of 1,385 probes
14	Claude 4.5 Opus	Low	3%	44 of 1,385 probes
15	GPT 5.6 Luna	Low	2%	34 of 1,645 probes
16	GPT-5.6 Sol	Low	2%	34 of 1,645 probes
17	GPT-5.6 Terra	Low	2%	31 of 1,645 probes
18	GPT-5 nano	Low	2%	25 of 1,385 probes
19	Gemini 3.8 Flash	Low	1%	19 of 1,645 probes
20	GPT-6 Astra	Low	0%	7 of 1,645 probes
21	Grok 4.6	Low	0%	5 of 1,649 probes
22	Claude Sonnet 5.5	Low	0%	4 of 1,655 probes
23	Claude 5 Opus	Low	0%	3 of 1,645 probes
24	Gemini 3.7 Flash	Low	0%	3 of 1,649 probes
25	Grok 4.7	Low	0%	1 of 1,038 probes
26	Claude Fable 5	Low	0%	1 of 1,645 probes
27	Claude Fable 5.1	Low	0%	1 of 1,645 probes
28	Claude 5.5 Opus	Low	0%	1 of 1,655 probes
29	GPT-5.5	Low	0%	0 of 3 probes

Head-to-Head
LLM Comparison

Compare the security of any two AI models, side by side.

Model

Attack success rate over time

—

Attack Success Rate

—

Successful attacks

—

Distinct high ASR vulnerabilities (over 80% attack success across recent scans) each model is susceptible to, and the ones they share

only

shared

only

Human-Powered AI Security Scoring

Our security ratings come from real-world attacks discovered by a global community of security researchers, not synthetic or auto-generated benchmarks. Every probe is a human-discovered exploit, rigorously validated before it counts toward a score.

Attack success rate (ASR)

(Average across scanned models)

Attack Success Rate

Low Risk Moderate Risk High Risk Critical Risk

The average AI model fails 1 in 15 attacks, meaning attackers succeed 6.5% of the time across all scanned production models. A higher Attack Success Rate (ASR) means a model is more easily jailbroken and at greater security risk.

Models scanned: 29
Number of unique probes: 1,697
Avg successful attacks: 95.1
Active researchers: 2,819
Submissions rejected: 83%

LLM Security Rankings

Get monthly LLMrankings in your inbox

Comprehensive security rankings

Head-to-HeadLLM Comparison

Attack success rate over time

Human-Powered AI Security Scoring

Attack success rate (ASR)

How does yourLLM rank?

Get monthly LLM
rankings in your inbox

Head-to-Head
LLM Comparison

How does your
LLM rank?