What is MiniMax?
MiniMax is a company founded in early 2022 that develops its own multimodal foundation models. Its lineup includes large language models (MiniMax M3, M2.7, M2.5) with a 1M-token context and a novel sparse attention architecture, the MiniMax H3 video generation model, MiniMax Speech 2.8 for text-to-speech and voice cloning, and MiniMax Music 3.0 for music generation. The models understand and generate text, audio, images, video and music.
On top of these models MiniMax offers products: MiniMax Code, a coding harness with Coding and Work modes for working with code tools; MiniMax Design, an agent platform for commercial content such as ads, e-commerce assets and campaigns; Audio; Talkie; and an API for developers and enterprises. The site also lists a Plan subscription and open-weight releases for some models.
Key features
- Frontier coding and agentic model with 1M context (M3)
- Video generation (MiniMax H3)
- Text-to-speech with voice design and voice cloning (Speech 2.8)
- Music generation with open weights (Music 3.0)
- Agent-driven workflows for commercial content creation (MiniMax Design)
- Coding and Work modes in the MiniMax Code desktop app
Who is it for
- developerzy i zespoły programistyczne
- firmy i zespoły marketingowe
- twórcy treści wideo i audio
- przedsiębiorstwa integrujące AI przez API
Pros and cons
- Models cover text, audio, image, video and music in one family
- 1M-token context window in M3
- Pay-as-you-go API pricing with a 50% discount on some models
- Priority tier costs 1.5x the standard price
- Some features are paid only, billed in tokens, characters or hours
MiniMax pricing
| Plan | Price | Includes |
|---|---|---|
| MiniMax-M2.7 | 0.3 USD input / 1.2 USD output per M tokens | Prompt caching read 0.06 USD / M tokens |
| MiniMax-M2.7-highspeed | 0.6 USD input / 2.4 USD output per M tokens | Prompt caching read 0.06 USD / M tokens |
| Speech recognition (ASR) | 0.38 USD per hour | Streaming, speaker diarization, srt/vtt export |
| T2A speech-2.8-hd (sync or async) | 100 USD per M characters | Async supports up to 1M characters per request |
| T2A speech-2.8-turbo (sync or async) | 60 USD per M characters | Same capabilities as hd, optimized for speed and cost |
| Voice Design | 3 USD per voice | Charged on first use for synthesis; in-API preview billed at 60 USD / M characters |
Prices copied from MiniMax pricing, checked October 12, 2026. Currency as shown by the vendor; taxes may apply.
Is MiniMax free?
The vendor does not state this clearly on the homepage or pricing page.
MiniMax alternatives
| Tool | Price from | Free plan | Category |
|---|---|---|---|
 Grok API | 0.02 USD | ? | APIs & models |
 Google AI Studio | 0,75 USD | ? | APIs & models |
 Replicate | 0.025 USD | trial | APIs & models |
 ElevenLabs | 6 USD/mo | yes | Voice & text to speech |
 OpenAI Platform | 0,10 USD | ? | APIs & models |
 DeepAI | 9.99 USD/mo | yes | APIs & models |
FAQ
What is MiniMax used for?
Multimodal AI models (LLM, video, speech, music) and AI-native products such as MiniMax Code, Design and Talkie, for developers, creators and enterprises.
Is MiniMax free?
The vendor does not state this clearly on the homepage or pricing page.
How much does MiniMax cost?
The cheapest paid plan is 0.3 USD input / 1.2 USD output per M tokens (vendor pricing, October 12, 2026).
What are the best MiniMax alternatives?
Closest alternatives: Grok API, Google AI Studio, Replicate, ElevenLabs, OpenAI Platform.
Data checked October 12, 2026 on the vendor website (minimax.io).