DeepThreat Benchmark Results
Beating Industry Leaders
82.6% detection rate on EVMbench — outperforming GPT-5.3-Codex and commercial tools.
Zero API costs. Local AI reasoning. 44 specialized scanners.
Performance Metrics
Competitive Comparison
| Tool | Detection Rate | Scanners | AI Reasoning | Cost | Speed |
|---|---|---|---|---|---|
| DeepThreatBEST | 82.6% | 44 | ✓ Yes | $0 | < 5 min |
| Slither | ~65% | ~90 | ✗ No | $0 | < 1 min |
| Aderyn | ~45% | ~40 | ✗ No | $0 | < 2 min |
| SolidityScan | ~70% | Unknown | ✓ Yes | $$$ | 5-10 min |
| GPT-5.3-Codex | 72.2% | N/A | ✓ Yes | $$$ | Variable |
44 Specialized Scanners
Testing Methodology
EVMbench Test Suite
Benchmarked against EVMbench, an industry-standard vulnerability detection test suite containing real-world exploit patterns from past DeFi hacks. Tests include reentrancy, access control, oracle manipulation, integer overflow, and economic attack vectors.
VulnLLM-R Integration
DeepThreat uses VulnLLM-R for local AI-powered vulnerability reasoning, eliminating API costs while providing deep semantic analysis of code patterns. This hybrid approach combines static analysis with LLM reasoning for superior detection accuracy.
Continuous Testing
850 out of 856 automated tests passing (99.3% pass rate). Test suite includes unit tests for individual scanners, integration tests for multi-scanner correlation, and regression tests against known CVEs and past exploits.
See the Difference For Yourself
Try DeepThreat's scanner suite on your smart contracts. Free. Local. No API keys required.