Official pricing and limits
| Field | Value |
|---|---|
| Model ID | MiniMax-M3 |
| Standard input <=512K / 1M | $0.30 |
| Standard output <=512K / 1M | $1.20 |
| Cache read <=512K / 1M | $0.06 |
| Standard input >512K / 1M | $0.60 |
| Standard output >512K / 1M | $2.40 |
| Priority multiplier | 1.5x standard |
| Context window | 1,000,000 tokens |
Where it fits
MiniMax M3 is a low-rate candidate for coding, tool-use, and multimodal agent experiments. Check how often your prompts cross 512K first, because the published input and output rates double above that boundary.
When to skip it
Skip it if your stack depends on provider-native enterprise controls from OpenAI, Anthropic, or Google, or if you need official max-output and benchmark tables before approval.