what's everyone's low-cost, high-throughput model these days for bulk processing? I was using gpt-5-mini but its content_filter is now rejecting more than half my API requests

Gemini-2.0-flash

GPT 4.1 nano