minimax Open weights Reasoning

MiniMax M2.5 pricing & benchmarks

Released Feb 12, 2026 · served by 14 providers

Input / 1M tokens$0.150
Output / 1M tokens$0.900
Context window205K
Cached input$0.050
1M in + 300K out$0.42

What MiniMax M2.5 is

MiniMax-M2.5 is a SOTA large language model designed for real-world productivity. Trained in a diverse range of complex real-world digital working environments, M2.5 builds upon the coding expertise of M2.1...

Providers and prices for MiniMax M2.5

The same model costs different amounts depending on who serves it. Prices are per 1M tokens, cheapest first. Uptime and throughput are OpenRouter's rolling measurements, not vendor claims.

ProviderInputOutputContext QuantUptime 24hThroughput
Inceptroncheapest $0.150 $0.900 197K fp8 99.9%
DigitalOcean $0.225 $0.900 66K 98.9%
Venice $0.270 $0.950 198K 94.8%
Mara $0.240 $0.960 197K 90.4%
StreamLake $0.270 $1.08 200K 99.1%
Chutes $0.150 $1.20 197K fp8 98.7%
AtlasCloud $0.295 $1.20 197K fp8 99.2%
Friendli $0.300 $1.20 197K 99.7%
Novita $0.300 $1.20 205K fp8 99.5%
SiliconFlow $0.300 $1.20 197K fp8 99.0%
Parasail $0.300 $1.20 197K fp8 99.8%
CoreWeave $0.300 $1.20 197K fp8 97.7%
Phala $0.200 $1.38 197K 97.5%
Minimax $0.600 $2.40 205K fp8 99.3%

MiniMax M2.5 benchmark results

Design Arena head-to-head results, as reported through the OpenRouter model API.

CategoryArenaRankEloWin rate
svg models #29 1200 54.5%
website models #35 1247 57.5%
codecategories models #36 1237 56.8%
3d models #37 1227 57.6%
gamedev models #39 1232 55.5%
uicomponent models #44 1209 53.4%
dataviz models #48 1201 51.2%

FAQ

How much does the MiniMax M2.5 API cost?

$0.150 per 1M input tokens and $0.900 per 1M output tokens. A typical workload of 1M input + 300K output tokens costs about $0.42.

Is MiniMax M2.5 free?

No. It is a paid model starting at $0.150 per 1M input tokens, though some providers offer trial credits.

Which provider is cheapest for MiniMax M2.5?

Inceptron at $0.150 input / $0.900 output per 1M tokens (fp8 quantization).

Can I self-host MiniMax M2.5?

Yes — weights are published as MiniMaxAI/MiniMax-M2.5 on Hugging Face, so you can run it on your own hardware.