Text M2.7 Model
MiniMax's flagship LLM optimized for conversational depth and logical consistency. Supports advanced **Prompt Caching** (Write/Read) to reduce latency and costs for massive context chains.
Hailuo Video 2.3
Next-gen video generation supporting up to **1080P resolution** and 10s duration. Features a "Fast" mode for rapid prototyping at reduced credit costs.
High-Def Audio (T2A)
State-of-the-art text-to-audio synthesis with high-fidelity outputs ($100 per 1M characters). Ideal for audiobooks, podcasts, and automated localized voiceovers.
Creative Music Engine
Specialized Music-2.6 engine generating full songs up to 5 minutes. Includes a separate lyrics generator tool for complete creative orchestration.
What is the Prompt Caching 'Read' vs 'Write' cost?
When you first send a large context (like a book or codebase), you pay a 'Write' cost to store it in cache. Subsequent requests reusing that context pay a significantly lower 'Read' cost (up to 80% discount) instead of full input rates.
How accurate is the token to character ratio?
For Chinese characters, 1,000 tokens is approximately 1,600 characters. For English, the ratio is closer to 1,000 tokens per 750 words. The calculator automates these conversions based on 2026 data.
Are the Hailuo Video generations free for testing?
Generally no, but MiniMax often provides trial credits. Fast mode (768P, 6s) is the most efficient starting point at $0.19 per clip.
Does Voice Cloning support cross-language use?
Yes, the Rapid Voice Cloning model ($1.50 per voice) can clone a voice from as little as 5 seconds of audio and apply it to any of the supported 20+ languages.
