What is MiniMax M2.5?
MiniMax's open-weight coding and agent model, at $0.30 per million input tokens.
Who is MiniMax M2.5 best for?
MiniMax M2.5 is best for developers running coding agents on a tight budget.
Low-cost coding agents
MiniMax reports 80.2% on SWE-Bench Verified, and output costs $1.20 per million tokens on its API.
Multilingual codebases
It was trained on more than 10 programming languages, including Go, Rust, Kotlin and TypeScript, across over 200,000 real environments.
Search and tool use
MiniMax reports 76.3% on BrowseComp with context management, for agents that browse and call tools.
Self-hosting
The weights are on Hugging Face with deployment guides for vLLM, SGLang and Transformers.
How much does MiniMax M2.5 cost?
MiniMax M2.5's weights are free to download, and on MiniMax's API it costs $0.30 per million input tokens and $1.20 per million output tokens, or $0.60 / $2.40 for the faster highspeed version.
- Weights on Hugging Face
- Commercial use allowed under the MiniMax Model License
- Guides for vLLM, SGLang and Transformers
- Cache reads $0.03, cache writes $0.375 per million tokens
- Paid from your account balance
- Listed by MiniMax as a legacy model
- Same results as M2.5, faster output
- Cache reads $0.03 per million tokens
MiniMax now lists M2.5 under legacy models on its API; MiniMax M2.7 costs the same per token and MiniMax M3 is its current flagship. The licence lets you use, change and redistribute the weights royalty-free as long as you keep MiniMax's notice and follow its prohibited-uses policy, which rules out military use among others. MiniMax estimates 1,000 tokens at about 750 English words. Full pricing on platform.minimax.io. We checked these prices on .
How do readers rate MiniMax M2.5?
No one has rated MiniMax M2.5 yet. Be the first.
Sign in to rate. One rating per account.
What are the best MiniMax M2.5 alternatives?
Three tools on Recaply do MiniMax M2.5's job. DeepSeek-V3.2 and Step-3.5-Flash are open source, so you can run them on your own server.
DeepSeek's open, MIT-licensed reasoning model for agents, with thinking during tool calls.
Moonshot's open multimodal model for visual coding and agent swarms of up to 100 sub-agents.
StepFun's 196B MoE with a 256K context for coding and agents.
Questions about MiniMax M2.5
What is MiniMax M2.5?
MiniMax's open-weight coding and agent model, at $0.30 per million input tokens.
Who is MiniMax M2.5 for?
MiniMax M2.5 is best for developers running coding agents on a tight budget.
Is MiniMax M2.5 free?
Yes. MiniMax M2.5 has a free Open weights plan: weights on Hugging Face.
How much does MiniMax M2.5 cost?
MiniMax M2.5's paid plans are Pay as you go at $0.30 / $1.20 per million input / output tokens and M2.5-highspeed at $0.60 / $2.40 per million input / output tokens.
Is there an open source alternative to MiniMax M2.5?
Yes. DeepSeek-V3.2 and Step-3.5-Flash are open source tools that do the same job, so you can run your own.
What are the best MiniMax M2.5 alternatives?
The closest alternatives on Recaply are DeepSeek-V3.2, Kimi K2.5 and Step-3.5-Flash.
Where can I try MiniMax M2.5?
On its official website, minimax.io.
Is MiniMax M2.5 open source?
Its weights are public on Hugging Face under the MiniMax Model License, which allows royalty-free use, changes and redistribution, including commercial use. It is not a standard open-source licence: you must keep MiniMax's notice and follow a prohibited-uses policy.







