Short answer
MiniMax M2 is an open-weight model MiniMax announced on October 27, 2025, for coding and agent workflows. Its official repository describes it as a mixture-of-experts model with 230 billion total parameters and 10 billion active parameters. For a new production integration, don’t assume MiniMax’s hosted API currently offers M2: the API model list identifies M3.1-Flash and M3 as available language models, while the reviewed invocation guide documents supported models rather than confirming an M2 access path. The reviewed pricing overview also does not show an M2 rate. (MiniMax announcement; M2 repository; API model list; pricing overview)
Selection takeaway: M2 may be worth investigating if you specifically need its released weights and can evaluate deployment in your own environment. If you need a MiniMax-hosted API for a new application, consider models currently listed in the API documentation—M3.1-Flash or M3—or confirm M2 availability with MiniMax before building around it. The pricing overview lists M3.1-Flash at $0.15 per million input tokens and $0.60 per million output tokens; that is an alternative’s price, not M2 pricing. (API model list; pricing overview)
What MiniMax M2 is
MiniMax announced M2 on October 27, 2025, describing it as a model for agents and code. Its official GitHub repository describes a 230-billion-parameter MoE model with 10 billion active parameters. These details identify the model and its intended workload; they do not establish how well it will perform on your application. (MiniMax announcement; M2 repository)
MiniMax’s announcement positions M2 for coding and agent use. Treat that as intended-use guidance, not a guarantee of accuracy, reliability, latency, or cost for a particular workload. The reviewed materials do not provide enough task-specific evaluation evidence here to make a performance comparison. (MiniMax announcement)
Is MiniMax M2 available through the production API?
The reviewed current official API documentation does not list M2 among its available language models. MiniMax’s model list names M3.1-Flash and M3, and its invocation guide documents supported models. The reviewed pricing overview lists rates for current models but no M2 price. Based on these documents, M2’s current hosted API availability and purchase terms cannot be confirmed. (API model list; invocation guide; pricing overview)
That documentation gap is not proof that M2 is unavailable through every access route. But the reviewed materials do not establish a current M2 model ID or request path for a hosted production API. Confirm the exact model ID and access terms with MiniMax before implementation; do not substitute M2.5 or M2.1 details, because they are different model versions. (API model list; invocation guide)
How to decide whether M2 fits
- Consider investigating M2 if your work centers on coding or agent workflows and you can evaluate the released model in your intended environment. Check the applicable model license and deployment requirements before commercial use. (M2 repository)
- Prefer a currently listed API model if you need MiniMax-hosted production access, documented invocation instructions, and a published current API price. M3.1-Flash and M3 are listed by MiniMax as available; the documented M3.1-Flash rate is $0.15/M input tokens and $0.60/M output tokens. (API model list; pricing overview)
- Don’t choose M2 based only on a “cheaper than competitors” claim. The reviewed current M2 materials do not establish an applicable M2 API rate or comparable, current cost-per-task results. Measure successful-task rate, retries, latency, and hosting costs for your workload before making a cost decision.
Bottom line
MiniMax M2 is an officially announced open-weight coding and agent model, distinct from M2.5 and M2.1. Its public model materials make it a candidate for evaluation where using released weights is suitable. But the reviewed current MiniMax API documentation lists M3.1-Flash and M3, not M2, and the reviewed pricing overview provides no M2 rate. For new hosted production use, start with currently listed API models or confirm M2’s distribution route, license, and operating costs with MiniMax before committing. (MiniMax announcement; M2 repository; API model list; pricing overview)
Sources
MiniMax M2 & Agent: Ingenious in Simplicity