Global AI Token Comparison Hub
CN
Back to all plans

MiniMax

M3 multimodal

M3 flagship — credits-based unified quota across all open-platform models

Token PlanToken Plan
SubscriptionToken API
Official

Core models

MiniMax-M3MiniMax-M2.7MiniMax-M2.7-highspeedSpeech-2.8MiniMax Hailuo 2.3image-01Music-2.6
MiniMax-M3

Current flagship multimodal model with native multimodal support and 1M context for frontier coding and complex agents—shares Token Plan credits with M2.7, deducted at API list price.

MiniMax-M2.7

Mainline language model for high-frequency text and coding—shares the same plan quota with M3 for everyday agent and programming workflows.

MiniMax-M2.7-highspeed

High-speed variant matching M2.7 quality with faster responses—suited to latency-sensitive high-concurrency and real-time assistant scenarios.

Speech-2.8

Speech-2.8 (HD/Turbo) supports sync and async long-text synthesis; speech resources are covered by Token Plan and share the credit pool with text models.

MiniMax Hailuo 2.3

Video generation (including Hailuo 2.3 Fast)—Ultra includes 5 daily videos; video usage deducts from the credit pool at list price.

See the official site for more models

Additional core model names still appear above, with full details on the latest official page.

Open official site

Plan details

Plus

Starter
Price
¥49/mo
3-4 agents · 5-hour & weekly windows
Official
Usage
At ¥49/mo, Plus targets light personal development and everyday trial as the entry public tier; M3 and M2.7 share one plan credit pool with access to all open-platform models.
Models
Typical agent use is about 3-4 runs/month within 5-hour and weekly quota windows—good for trying agents, coding, and multimodal mixes rather than all-day heavy production.
Highlights
Usage deducts credits at each model’s API list price; text, image, speech, music, and more share one pool, shown as a unified progress bar in the console.
If you mainly want to validate MiniMax as a multimodal agent workbench, Plus is the natural entry; when quota runs out, buy credits or switch to pay-as-you-go API.
Best for
Entry users, small teams, and light-to-medium agent users exploring multimodal workflows

Max

RecommendedRecommended
Price
¥119/mo
4-5 agents · 5-hour & weekly windows
Official
Usage
At ¥119/mo, Max targets high-frequency coding agents and multimodal calls as the most balanced primary tier, with full M3 and M2.7 access across all open-platform models.
Models
Typical agent use is about 4-5 runs/month with more headroom in 5-hour and weekly windows than Plus—suited to making Claude Code, Cursor, OpenClaw, and similar tools part of daily primary workflows.
Highlights
The cross-modal credit pool flexibly covers text, image, speech, and music; if you combine synthesis, agent collaboration, and content generation, this tier is usually steadier than Plus.
When window limits are hit, purchased credits can auto-cover overage, or you can upgrade to Ultra or switch to pay-as-you-go API keys.
Best for
Advanced developers, frequent multimodal users, and people treating agents as part of a regular toolchain

Ultra

Power
Price
¥469/mo
6-7 agents · ~7.1B tokens/mo · 5 videos/day
Official
Usage
At ¥469/mo, Ultra targets heavy agent workflows and longer sessions with roughly 7.1B tokens/month—the highest quota among public subscription tiers.
Models
Typical agent use is about 6-7 runs/month with 5 daily video generations; M3 long-horizon tasks consume more, making this tier better for treating MiniMax as a production system.
Highlights
Text, image, speech, music, and video (per tier perks) all deduct from one credit pool—suited to long parallel agent runs and multimodal content production.
If your workflow continuously mixes agents, multimodal generation, and video, Ultra is where the gap from mid tiers becomes truly meaningful.
Best for
Heavy multimodal users, complex content teams, and people running sustained high-frequency agent workflows

Quota

Product type
Token Plan
Official product line
Token Plan
5-hour limit
5-hour rolling window quota
Weekly limit
Weekly window quota
Monthly limit
—

Notes

  • Plan quota and purchased credits cover the same resources; plan quota is used first, then credits. Subscription keys and pay-as-you-go API keys are not interchangeable.
  • Credit packs: ¥30/4,285 credits, ¥150/21,430 credits, ¥500/71,435 credits, valid 365 days from purchase.
  • Agent counts (3-4 / 4-5 / 6-7) are estimates from typical task token usage; actual capacity varies by model and modality—see the console progress bar.
  • Public pricing docs list monthly pay only (Plus ¥49, Max ¥119, Ultra ¥469); annual offers if any are on the subscribe page—we do not show unverified annual prices.
  • Plus/Max have no published token cap—use agent estimates (3-4 / 4-5) and 5-hour/weekly windows; only Ultra cites ~7.1B tokens/month in the migration notice.
  • Speech and video resource packs are separate prepaid products (see pricing-speech / pricing-video), bought apart from Token Plan; in-plan speech/video usage still deducts credits at API list prices.

Supported coding tools

Claude CodeCursorClineRoo CodeKilo CodeCodex CLIOpenCodeDroidTRAEGrok CLI

Pricing and model data sourced from official vendor websites

FAQ

General·7
General
7 条