MiniMax: MiniMax M2

MiniMax M2 is a text generation model from MiniMax, released on 23 October 2025. It is a compact, high efficiency large language model with 10 billion activated parameters out of 230 billion total, and MiniMax describes it as optimized for end to end coding and agentic workflows. It sits in the text category, so it is aimed at language tasks rather than image, audio or video generation.

The model takes text as input and returns text as output. It has a context window of 204,800 tokens, which is large enough to hold substantial source files, long documents or extended conversation histories in a single request. The maximum output is 176,947 tokens, so it can produce very long completions when a task requires them. MiniMax positions it as delivering near frontier intelligence across general reasoning, alongside its focus on coding and agent style work.

Those capabilities point to a few practical uses. The coding focus makes it a reasonable fit for code generation, refactoring, debugging and working across multi file repositories that need to be read together within the context window. The agentic emphasis suits tool calling loops, where the model decides which function to invoke and incorporates the result into its next step. The long context and long output limits also help with summarising or transforming large documents, and with reasoning heavy tasks where the model benefits from working through a problem before answering.

You can call it through api14's OpenAI compatible API using the model id "minimax/minimax-m2". The supported request parameters are frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs and top_p. Tool use is available through tools and tool_choice, structured outputs through response_format and structured_outputs, and reasoning behaviour can be influenced with reasoning and include_reasoning. Sampling can be tuned with temperature, top_p, top_k, frequency_penalty, presence_penalty and repetition_penalty, while seed, stop, logprobs and top_logprobs give you control over reproducibility and token level output.

Since the model accepts only text input, tasks involving images, audio or other modalities need a different model. If your workload does not involve coding, agentic tool use or long context reasoning, a smaller or more general text model may be simpler and sufficient.