These figures come from the per-token board. They are prices per million tokens, not per request.
Inference provider cash prices · Lowest 0.55125 · median 0.55125 · highest 0.55125 · 1 quotes USD / per million tokens
Relay CNY credit prices · Lowest 0.25 · median 3.3075 · highest 21.058 · 22 quotes CNY / per million tokens, site credit
Relay USD credit prices · Lowest 0.18375 · median 1.8375 · highest 6.615 · 43 quotes USD / per million tokens, site credit
Each quote is blended into one price by this workload: ordinary input 75% · cache read 0% · output 25% · cache write 0%. Only prices in effect now are counted.
Uses the standard tier (the tier that contains 8K input tokens, base price). Full rules are in Details.
All tiered prices: 56,110 prices from 1,093 sites (about 1,004 companies). 50,064 are ranked, from 1,079 sites (about 990 companies); 6,046 more are listed separately and not ranked. Prices collected up to 2026-11-15.
This model has 25 tiered prices; 20 can be compared directly. Not comparable yet: display unit not confirmed 3, new price under review; showing the previous one 2
11 recent · 0 stale (older than 10 days, or not updated by this refresh) · 0 unverified timestamps. Minimum and median only count recent paid prices; older prices stay in the table.
For tiered models (e.g. longer context costs more) we use the standard tier: the tier that contains 8K input tokens, normal service tier, base price without promotions. Prices that change with time of day or request options are listed separately with the reason. See “Details” for the full rules.
Prices are converted to site credits using each site’s published ratio and display multiplier. What you actually pay depends on its top-up rate. Several sites run by the same company take one place in the ranking: its cheapest plan, with the other sites listed under it. We judge “same company” by identical prices, status pages or homepage code; this is not a legal finding. To see every site’s prices, choose “All plans” or filter by provider.
These manufacturer statements do not verify relay backends, availability, GPU ownership or cash eligibility. Aliases are never merged automatically.
MiniMax-M2.5
Manufacturer-declared status
Retired (retirement date 2026-11-18 has passed)
New-user access cutoff
2026-10-19
Model retirement date
2026-11-18
Existing pay-as-you-go users can continue until retirement; M Plan / Token Plan compatible names route to M3.1 Flash Preview from 2026-10-19. This route change is plan-specific, not a universal identity alias.
The vendor has stopped this model, or its stated retirement date has passed. This does not mean a paid test call failed.
Complete declarations and historical observations
{
"author": "MiniMax",
"availability": "EXISTING_PAYG_USERS_ONLY_PENDING_RETIREMENT",
"decision": "ADD_GROUP",
"documentedAliases": [],
"entityId": "minimax/minimax-m2.5",
"evidenceKeys": [
"minimax-models",
"minimax-sdk"
],
"newUserAccessStopsOn": "2026-10-19",
"note": "Existing pay-as-you-go users can continue until retirement; M Plan / Token Plan compatible names route to M3.1 Flash Preview from 2026-10-19. This route change is plan-specific, not a universal identity alias.",
"officialId": "MiniMax-M2.5",
"qualificationEffect": "NONE_NAME_DECLARATION_ONLY",
"retirementDate": "2026-11-18",
"sourceUrls": [
"https://platform.minimax.cn/docs/guides/models-intro",
"https://platform.minimax.cn/docs/api-reference/text-anthropic-api"
],
"verifiedOn": "2026-10-19",
"w42ObservedAt": [
"2026-10-17T04:01:32.471Z"
],
"w42OfficialPriceNames": [
"MiniMax-M2.5"
],
"w42OfficialPriceRows": 1
}