MiniMax M2.5
MiniMax M2.5Rankingin thinking mode?
Alternatives?
CheaperMiniMax M3.1 Flashthinking×2.1-50%$0.22Complete model response must be appended
BetterGPT-6 Lunalow+56%-62%$0.005Function calling only with none · Chat Completions+3nOpenNex N2 Pro+33%+90%$0.50 / $2.50Memory (4-bit): 137 GB → 238 GBCheaperGPT-5.6 Lunamax+63%-14%$0.18
CheaperClaude Haiku 5.5max+39%-62%$0.21Remove sampling parameters+5MiniMax M3.1 Flashthinking×2.1-50%$0.22Complete model response must be appendedGPT-6 Lunaxhigh+69%-62%$0.042Function calling only with none · Chat Completions+3GPT-5.6 Lunamax+63%-14%$0.18Claude Haiku 5.5max+39%-62%$0.21Remove sampling parameters+5RBeam 1.0 501B A23BthinkingOpen+26%-17%$0.57Memory (4-bit): 137 GB → 301 GBGPT-6 Lunalow+56%-62%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow+51%-14%$0.01Claude Haiku 5.5low+27%-62%$0.024Remove sampling parameters+5
Gemini 3.1 Flash Litehigh+39%+7%$0.06
Hy3thinkingOpen+22%-56%$0.072Memory (4-bit): 137 GB → 179 GBHy3thinkingOpen+22%-56%$0.072Memory (4-bit): 137 GB → 179 GB
Qwen3.5 397B A17BthinkingOpen+18%+157%$0.47Memory (4-bit): 137 GB → 242 GBMiniMax M3.1 Flashthinking×2.1—$0.22Complete model response must be appendedMiniMax M2.7thinkingOpensame—$0.30 / $1.20MiniMax M2thinkingOpen-11%—$0.30 / $1.20MiniMax M1 80kOpen-14%—$0.55 / $2.20Memory (4-bit): 137 GB → 274 GBMiniMax M2.1thinkingOpen-19%—$0.30 / $1.20MiniMax M3thinkingOpen-36%—$0.51Memory (4-bit): 137 GB → 256 GB+2MiniMax M1 40kOpenn/a——Memory (4-bit): 137 GB → 274 GBMiniMax M1unstatedOpenn/a—$0.40 / $2.20Memory (4-bit): 137 GB → 274 GB+1All MiniMax models →Run it yourself
estimateHistory
6 versionsBenchmarks
all categoriesFAQ
How much does MiniMax M2.5 cost?›
$0.30 per 1M input tokens and $1.20 per 1M output tokens; cached input $0.03 per 1M; cheapest via Azure AI Foundry at $0.033 / $1.32.
Is MiniMax M2.5 free to use?›
There is no free API; the cheapest paid price is $0.033 per 1M input tokens. The weights are open, so you can run it yourself.
How good is MiniMax M2.5, and what is it best at?›
#122 of 334 in Chat & Knowledge, #74 of 97 in Coding, #18 of 242 in Agents & Tool Use, #98 of 195 in Writing & Language. Strongest at: Tool calling, Instructions.
Is MiniMax M2.5 better than MiniMax M2.1?›
On AA Omniscience · Accuracy, MiniMax M2.5 scores 26.2 vs 21.2 — higher.
How can I use MiniMax M2.5?›
Through the API of AWS Bedrock, AtlasCloud, Azure AI Foundry, DeepInfra, DigitalOcean and 4 more.
What is the context window of MiniMax M2.5?›
205K tokens, with up to 131K output tokens.
When was MiniMax M2.5 released? Is there a newer version?›
Released 2026-02-12. It has been superseded by MiniMax M2.7.
Is MiniMax M2.5 open source?›
Yes, the weights are open (modified-mit licence).
Can I run MiniMax M2.5 locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can MiniMax M2.5 read images?›
No, it accepts text only.
What is MiniMax M2.5?›
MiniMax M2.5 is an open-weights language model from MiniMax, released Feb 2026, the successor to MiniMax M2.1.
What are the best alternatives to MiniMax M2.5?›
MiniMax M3.1 Flash (cheaper), GPT-6 Luna (better), Nex N2 Pro (open weights).