Tools
Groq
Groq is an LLM inference provider offering ultra-low-latency (~200ms first-token) on open-source models (Llama, Mixtral, Gemma). Different from the Groq cloud — same company, name reuse intended.
More detail
Aiprosol's 10 AI agents run on Groq for cost + speed reasons. Pricing: $0.05-$0.59 per million tokens (vs Claude $3-$15). At our usage levels, monthly cost is <$5. Trade-off: Groq's models are open-source (Llama 3.3 70B is the strongest), less accurate than frontier models like Claude on judgement-heavy tasks. We use Groq for the bulk-work agents (data, ops) and reserve Claude for customer-facing chat.