MiniMax-M2.5-fast was offered as a text-to-text language model on the Nebius Token Factory serverless catalog, where it carried the MiniMaxAI/MiniMax-M2.5-fast identifier and supported the kinds of inference workloads typical of fast-tier chat models, including reasoning, tool calling, structured output, and temperature control. The catalog marked the entry as open weights, indicating that the model parameters were distributed for self-hosting or local deployment in addition to the hosted inference surface. For most of its lifespan it sat alongside other open-weight chat and reasoning models that Nebius exposed through its API gateway, giving developers a familiar chat-completion interface with tunable sampling.

The model is now part of Nebius Token Factory's June 22, 2026 deprecation cohort: the official deprecation page lists MiniMaxAI/MiniMax-M2.5-fast among the affected endpoints whose APIs and UI are being disabled, alongside several other chat, reasoning, and mixture-of-experts models in the same batch. The Sulat catalog entry reflects this status with a deprecated badge and links users to the Nebius deprecation guidance. In practical terms, developers still relying on this model should treat it as end-of-life on the Nebius platform and plan migrations to other supported endpoints or to dedicated deployments that Nebius offers for production workloads.