MiniMax M

Overview

This is a model family overview. For version-specific details, see the individual model entries linked below.

MiniMax M is MiniMax’s flagship model family for high-capability reasoning, coding, multimodal understanding, and long-context tasks. The line began as proprietary hosted language models, but M3 adds native image and video input, a 1M-token context budget, and official open weights alongside MiniMax’s managed API.

Current Latest

MiniMax M3 is the current flagship. It was released on June 1, 2026 as a 1M-context native multimodal model for coding and long-horizon agents. MiniMax M2.7 remains an active older text-only API tier, with M2.7-highspeed offering the same model capability at higher throughput and higher token prices.

Strengths

• Up to 1M total-token context in the current M3 generation.
• Competitive coding and reasoning quality in recent M-series releases.
• Strong fit for bilingual Chinese-English workloads.
• Native image and video input in M3.
• Flexible managed-API and open-weight deployment paths in the latest generation.
• Production-oriented emphasis on coding, search, tools, computer use, and office deliverables.

When to Choose MiniMax M

Choose MiniMax M when you want a lower-cost frontier-style route for coding, multimodal analysis, office deliverables, and longer agent workflows. M3 is the current default for new evaluations; M2.7 remains useful for pinned text-only integrations.

Access

• MiniMax platform APIs and managed services
• Coding Plan subscriptions for developer-heavy usage
• Compatible Anthropic-style API integrations in MiniMax docs
• Official MiniMax M3 open weights on Hugging Face under the source-available MiniMax Community License