FlowCay logo
FlowCay
MiniMax

MiniMax

AI Agent
7/12/2025

Overview

MiniMax multimodal AI platform

Details

Overview

MiniMax (上海稀宇科技, founded 2021) is a Chinese AI company offering a full multimodal model family: the M-series language models, Hailuo AI video generation, speech synthesis, music and image models. It sells through two channels — pay-as-you-go API pricing for enterprises and monthly Subscription (Token) Plans for individuals and small teams — plus consumer products like the Hailuo AI video app. MiniMax is known for aggressively cheap API pricing and is one of the strongest text-to-video models available.

Key Features

  • Language models (M-series): MiniMax-M3 is the flagship with a 1,000,000-token context window built for agentic reasoning, tool use, coding and long-context tasks. M2.7, M2.5, M2.1 and M2 offer 204,800-token context (~60 tps output, ~100 tps on "highspeed" variants). APIs are accessible via HTTP, the Anthropic SDK (recommended) and the OpenAI SDK.
  • Hailuo video generation: Hailuo 2.3 (text-to-video & image-to-video with strong instruction following and physics), Hailuo 2.3 Fast (value tier), Hailuo 02 (512p–1080p), and MiniMax-H3 — the next-gen open multimodal video model accepting text/image/first-last-frame/reference input at 768P–2K, 4–15s, 24fps.
  • Speech (TTS & voice cloning): speech-2.8-hd (ultra-realistic, sound tags) and speech-2.8-turbo (~40 languages on 2.6), rapid voice cloning ($1.50/voice) and voice design ($3/voice).
  • Music generation: music-3.0 (vocal songs from description + lyrics), music-2.6, music-cover (one-step and two-step cover generation with lyrics modification, style transfer) and music-2.0.
  • Image & vision: image-01 image generation ($0.0035/image) and API-vlm multimodal vision.
  • Agent & MCP: MiniMax Agent product, plus official MCP servers (Python & JavaScript) for speech synthesis, voice cloning, video and music generation.
  • Consumer apps: Hailuo AI (video generator), MiniMax conversational assistants and agent products with credit subscriptions.

Pricing

  • Pay-as-you-go API (per 1M tokens): entire M-series at a flat $0.30 input / $1.20 output (cache read $0.06). M3 above 512K input or Priority tier bills $0.60/$2.40 (Priority $0.45/$1.80). Resellers undercut further (e.g. OpenRouter lists M2.5 at $0.12/$0.48, M2.7 at $0.18/$0.72). TTS: speech-2.8-turbo $60/1M chars, speech-2.8-hd $100/1M chars. Hailuo video per clip: $0.19–0.33 (Fast), $0.28–0.56 (2.3/02). Music $0.15 per up-to-5-min track, lyrics $0.01/song (Music-3.0-free is free at 3 req/min). Image $0.0035/image.
  • Token Plan (monthly subscription): Starter $10, Plus $20, Max $50, Ultra $120/month — flat monthly quota (requests/5h window) across text, speech, image, music and video under one key; Highspeed variants annual-only from $400/yr. Annual billing gives ~2 months free.
  • Prepaid video packages: $1,000 (3,760 video points) to $6,000 (26,780 points) per month with 5–20% savings and 20–50 RPM.
  • Audio subscription: $5/month for recurring TTS workloads.
  • Hailuo AI consumer app: free trial up to Max $199.99/month (Standard $14.99, Pro $54.99, Master $94.99, Ultra $124.99); promotional pricing on Standard/Pro/Master at the time of writing.

Best For

Developers and startups that want cheap, high-quality multimodal APIs (especially the 1M-context M3 and Hailuo video), Chinese and global teams building agents with OpenAI/Anthropic-compatible calls, and creators using Hailuo AI for text-to-video.

FAQ

How much does the MiniMax API cost?

The M-series is billed at a flat $0.30 per 1M input and $1.20 per 1M output tokens (M3 over 512K input: $0.60/$2.40). Speech, video, music and image APIs bill on their own per-character / per-clip / per-track / per-image meters.

What is the Token Plan?

A monthly flat-rate subscription (Starter $10 to Ultra $120/mo) that bundles text, speech, image, music and video generation under one key and one bill, with quotas tracked over rolling 5-hour and weekly windows.

What is Hailuo?

Hailuo is MiniMax's text/video generation line (Hailuo 2.3, 2.3 Fast, Hailuo 02, H3), available both as the consumer Hailuo AI app and through the Video Generation API.

Does MiniMax offer an MCP server?

Yes — official MCP servers for Python and JavaScript cover speech synthesis, voice cloning, video generation and music generation.

Is MiniMax-M3 open source?

M3 is the closed-source flagship API model. MiniMax has open-sourced other models in the family (e.g. MiniMax-01 / Hailuo weights); check the model page for the latest open-weight releases.