Ask (cheapest). Rock-bottom-cost inference for high-volume, simple tasks. Powered by meta-llama/llama-3.3-70b-instruct. Max 600 output tokens.