← All providers
MiniMax pricing
verified 2026-09-02Shanghai-based lab with long-context open-weight models priced for high-volume workloads.
data residency & retention
from MiniMax's own policies · verified 2026-09-02 · some facts not publishedHeadquarters
🇨🇳 China (China)
Where inference runs
Not disclosed
The international platform (Singapore entity) discloses personal data is stored in a US datacenter, but inference compute location is not disclosed and transfers to unnamed affiliates (Shanghai parent) are permitted.
Region pinning
No
Only one international base URL is documented, with no regional endpoints or residency options.
Prompt retention
Retained
Input and output are retained as long as necessary or permitted by law — no fixed duration disclosed.
Trains on API data
Not disclosed
Terms grant broad rights to use input and generated content to improve services, with no opt-out disclosed — but neither an explicit training statement nor a no-training commitment.
Zero data retention
Not offered
No ZDR, ephemeral mode, DPA, or trust page found on any minimax.io property.
current models · $ per 1M tokens
| Model | Input /1M | Cached in | Output /1M | Context |
|---|---|---|---|---|
MiniMax M3 MiniMax M3 | $0.30 | $0.06 | $1.20 | 1M |
MiniMax M2.7 High-Speed MiniMax M2 | $0.60 | $0.06 | $2.40 | — |
MiniMax M2.7 MiniMax M2 | $0.30 | $0.06 | $1.20 | — |
MiniMax totals · 1M in + 1M out
MiniMax M3
$1.50
MiniMax M2.7
$1.50
MiniMax M2.7 High-Speed
$3.00
Cost of 1M input + 1M output tokens. Bar length is square-root scaled so cheap models stay visible next to premium ones.
legacy models · still purchasable
| Model | Input /1M | Cached in | Output /1M | Context |
|---|---|---|---|---|
MiniMax M2.5 MiniMax M2 | $0.30 | $0.03 | $1.20 | — |
MiniMax M2.1 MiniMax M2 | $0.30 | $0.03 | $1.20 | — |
MiniMax M2 MiniMax M2 | $0.30 | $0.03 | $1.20 | — |