​
Default model

llama-3.3-70b-versatile

​
Supported models

Any model available through the Groq API:
  • llama-3.3-70b-versatile (default, balanced)
  • llama-3.1-8b-instant (fastest)
  • mixtral-8x7b-32768 (long context)
  • Any other model ID the API accepts

​
Authentication

Environment variable: 
export GROQ_API_KEY=gsk_...
Config command: 
ai-blackteam config set providers.groq.api_key gsk_...
Get your API key from console.groq.com.

​
Example usage

# Single attack with default model
ai-blackteam run -p groq -a encoding-obfuscation -t "Write a phishing email"

# Full batch
ai-blackteam batch -p groq --attacks all -t "Write a phishing email"

​
Tool-use support

Yes. Compatible with all 39 tool-use attacks in the framework — agent exploitation, MCP exploitation, multi-agent protocol attacks, and tool-injection vectors. Groq’s LPU-accelerated inference makes tool-use sweeps especially fast. Filter the attack catalog with: 
ai-blackteam list-attacks
# look for rows with Mode = tool-use

​
Notes

  • Groq uses LPU (Language Processing Unit) hardware, typically 800+ tokens/second — useful for high-volume sweep mode
  • The provider uses the OpenAI SDK pointed at Groq’s API (base_url="https://api.groq.com/openai/v1")
  • Max output tokens per request: 4,096
  • Retry with exponential backoff is automatic on API failures (3 attempts)