Claude 3.5 Sonnet leads in AI code auditing benchmark, beating GPT-4o and Llama 3

A recent benchmark evaluated GPT-4o, Claude 3.5 Sonnet, and Llama 3 70B for their effectiveness in automated code auditing, specifically for detecting vulnerabilities in smart contracts. Claude 3.5 Sonnet emerged as the top performer in accuracy, correctly identifying all critical flaws and providing secure code patches with zero hallucinations. GPT-4o excelled in speed and adherence to structured JSON output, though it incorrectly flagged a non-existent vulnerability. Llama 3, when run locally, offered the fastest response times and privacy benefits but struggled with pure JSON output and exhibited moderate hallucinations. AI

IMPACT Claude 3.5 Sonnet shows superior accuracy for security-critical code auditing, while GPT-4o offers better structured output and speed for automated pipelines.

RANK_REASON Comparison of LLM performance on a specific technical task. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

• Anthropic
• Claude 3.5 Sonnet
• Common Vulnerability Scoring System
• GPT-4o
• HumanEval
• Llama 3
• Massive Multitask Language Understanding
• Meta
• Ollama
• OpenAI
• Solidity
• Vault

• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →