Guide
Choose MiniMax model, billing, agent-team and media-generation surfaces without combining unrelated subscriptions, credits or rights.
Published · Updated
Use M3 for model capability, API for production integration, Token Plan for supported coding tools, Mavis for long-running role-based agent work, MiniMax Design for orchestrated creative production, and dedicated media products for generation-specific workflows.
The guide uses official M3, Mavis Agent Team, MiniMax Design and Token Plan sources checked August 25, 2026.
M3 is positioned as an open-weight, native-multimodal model for coding and agents. Context limits, throughput and prices vary across MiniMax API and third-party providers. Compare the official model capability separately from each hosted offering.
Open-weight model identity, architecture and capability.
First-party metered serving and model-specific API terms.
API billing is usage-based. Token Plan is a subscription route for supported coding agents and tools. Even when subscriptions share credits with Agent features, verify included models, quotas, concurrency, renewal and enterprise terms before treating the plan as a production budget.
Application integration with usage and model-specific rates.
Subscription access for documented compatible coding workflows.
Mavis Agent Teams can split work across roles, run agents in parallel and preserve reusable experience. Adopt it when task decomposition and independent review justify added cost and control complexity; a single agent remains simpler for short interactive tasks.
Long-running work with role separation, parallel execution and review.
Prefer for short tasks where coordination overhead adds little value.
MiniMax Design orchestrates briefs, canvas workflows, skills, local assets and review checkpoints. Hailuo, Speech, Music and Image expose generation-specific products. Compare pricing units, output rights, likeness or voice consent, retention and moderation separately for each medium.
Agentic creative orchestration and local asset workflow.
Evaluate video, speech, music and image under medium-specific terms.