#Concurrency

8 Agents, One Local LLM — Ollama Concurrency Measured

I fired 8 agents at one local model expecting a speedup. Default Ollama serializes requests, so eight at once matched one. I benchmarked OLLAMA_NUM_PARALLEL.