Grok 4.20 API pricing

The three members are priced identically, and that is the useful fact here: $1.25 per million input tokens, $0.20 cached and $2.50 output over a 1,000,000-token window, for grok-4-20-0309-reasoning, grok-4-20-0309-non-reasoning and grok-4-20-multi-agent-0309 alike. A comparison table will show three identical rows, so cost cannot arbitrate. What varies is how many tokens each mode burns on the same task. The non-reasoning build answers directly and generates the least; the reasoning build emits a chain first, billed at the output rate, twice what input costs; the multi-agent build runs several passes and multiplies both meters. Choose on measured token counts and latency, then apply the $0.20 cache rate to whatever prefix repeats.

The rest of the range, and every rate xAI publishes: xAI API pricing