HighFlame

Trusted LLMs

OSS Model Supply Chain Security Leaderboard

Not all models are equally safe to deploy. This leaderboard provides supply chain security metrics for open-source LLMs across business domains so teams can evaluate trustworthiness before fine-tuning or production use. Models are scored using signals including artifact safety, tampering indicators, and behavioral risk.

Security posture may change over time as models are updated or redistributed.

Use it to:

Compare models beyond accuracy benchmarks
Assess deployment risk before fine-tuning
Decide which models to trust in production

Powered by Palisade

Comprehensive security validation for ML model artifacts

15+ Validators

Pickle RCE, backdoor detection, buffer overflow, tokenizer hijacking, and more

Model Signing

Cryptographic signatures with Sigstore/Cosign for artifact integrity verification

SLSA Provenance

Supply chain attestations following SLSA framework for build verification

SafeTensors & GGUF

Native validation for modern safe formats with integrity checks

Backdoor Detection

Statistical analysis for weight anomalies, baseline drift, and trojan patterns

Behavioral Analysis

Inference-time monitoring for prompt injection and tool call manipulation

MITRE ATLAS Mapping

Findings mapped to ML attack taxonomy for threat intelligence

Policy Enforcement

Configurable allow/deny/quarantine policies for CI/CD integration

Models Scanned

0

Safe Models

0

Avg Score

0

Critical Risk

0

Loading models...

Score Calculation

Security Score Formula

Score = (Critical × 100) + (High × 50) + (Medium × 10) + (Low × 2) + (Info × 1)

Lower scores are better. A score of 0 means no security issues were found.

Critical× 100 points
High× 50 points
Medium× 10 points
Low× 2 points
Info× 1 points

Example Calculations

1 critical finding= 100 points
2 high + 5 medium= 150 points
10 medium findings= 100 points
No findings= 0 points ✓

Risk Level Classification

Risk level is determined by the severity distribution of findings, not just the total score.

SafeNo significant findings
LowOnly low-severity issues
Medium1+ high OR 5+ medium findings
High3+ high severity findings
Critical1+ critical severity finding

Classification Logic

if critical ≥ 1 → Critical
elif high ≥ 3 → High
elif high ≥ 1 OR medium ≥ 5 → Medium
elif medium ≥ 1 OR low ≥ 5 → Low
elseSafe