MiniMax-M3 (EU)

MiniMax-M3 (EU) is positioned as a multimodal entry in the minimax family, engineered for long-context coding, perception tasks, and agent planning workflows. The EU-region variant is hosted as a gateway offering through Requesty, reflecting a generally available release with open weights that allow self-hosting or routing through alternative infrastructure. The model is designed to accept text, image, and video inputs while emitting text, which positions it for scenarios that combine visual understanding with extended document or code analysis. Its open-weight availability broadens deployment flexibility for teams that prefer local inference or custom serving arrangements over pure API consumption.

From a practical standpoint, the model combines reasoning and tool-calling capabilities with structured output support and temperature control, making it suitable for agentic pipelines and deterministic data extraction tasks. The wide context window enables processing of large codebases or lengthy multimodal transcripts in a single pass, useful for repository-level analysis or extended perception tasks. The Requesty gateway pricing—$0.40 per million input tokens and $2.00 per million output tokens, with a $0.10 cache-read rate—keeps the EU variant competitively positioned among open-weight multimodal options. Teams building coding assistants, document-understanding agents, or multimodal planning systems will find the combination of open weights, long context, and multimodal input particularly relevant for cost-sensitive or privacy-aware deployments.