Minimax Models

About Minimax

From public sources

MiniMax was founded in early 2022 and listed on the Hong Kong Stock Exchange in January 2026. Alongside its models, it runs its own consumer products.

Its language models are built for coding, agent workflows and office tasks, and several of its models are published with open weights on Hugging Face.

Developer: MiniMax
Founded: 2022
Headquarters: Shanghai, China
Website: minimax.io

Same API key for every model

Output modalities

MiniMax H3

Input: Text, Image, Audio, Video. Output: Video, Audio.

In – Out – Tiered /1M

MiniMax-M3

Input: Text, Image, Video. Output: Text.

In $0.60 Out $2.40 Cached $0.12 Tiered /1M

MiniMax-M2.7

Input: Text. Output: Text.

In $0.30 Out $1.20 Cached $0.06 /1M

MiniMax-M2.5

Input: Text. Output: Text.

In $0.30 Out $1.20 Cached $0.06 /1M

No models match your filters

Try loosening or removing some filters

Other providers

All 15 providers

OpenAIUnited States Output: Text, Image, Video, Audio. View provider GoogleUnited Kingdom Output: Text, Image, Video, Embeddings, Audio. View provider AlibabaChina Output: Text, Image, Video, Audio. View provider GLMChina Output: Text. View provider xAIUnited States Output: Text, Image. View provider AnthropicUnited States Output: Text. View provider DeepSeekChina Output: Text. View provider ByteDanceChina Output: Video, Audio. View provider CohereCanada Output: Text, Embeddings. View provider KlingChina Output: Video, Audio. View provider MoonshotChina Output: Text. View provider TencentChina Output: Text. View provider MetaUnited States Output: Text. View provider TypeSafeUnited States Output: Decisions. View provider

One endpoint, every Minimax model

Switch Minimax models by changing one model ID

Create a MixRoute API key, set the base URL, and call any Minimax model above with the same request format. Usage across all of them lands on one bill.

Create an API key Browse all models