MiniMax AI Review 2026: H3, Video, Speech, API Pricing, M Plan Guide
Chinese AI company with M3.1 Flash and M3 language models, the open-weight H3 video model, Speech 2.8 and Music 3.0, plus MiniMax Code and an API.
MiniMax AI: M3.1 Flash, H3 Video, Speech and Music Models
MiniMax is a Chinese general AI company that builds its own models for text, video, speech and music: the M3.1 Flash and M3 language models, the open-weight H3 multimodal and video model, Speech 2.8 and the open-weight Music 3.0. It turns them into products such as MiniMax Code, MiniMax Design, a general agent and the Hailuo AI video app, and sells API access to developers. MiniMax says it serves 300M+ individual users and 2M+ business clients and developers in 230+ countries and regions.
MiniMax sits alongside other Chinese AI providers covered on the site, including DeepSeek, Qwen and StepFun. For broader comparisons, see our guides to Chinese AI chatbots, free Chinese AI tools and Chinese AI video generators. You can also read the Hailuo AI review, compare Claude, or browse AI chatbots and assistants.
Quick Navigation
Quick Summary
Overall Rating 4.2/5 Best For Developers and creators who want MiniMax models for agents, coding, video, voice and music Ease of Use 4.0/5 Business Value 4.3/5
What Is MiniMax and Why Does It Matter?
MiniMax develops proprietary and open-weight foundation models across text and multimodal tasks. MiniMax M3.1 Flash, released in October 2026, is aimed at fast, low-cost agent work, and M3 is its larger language model. MiniMax H3, released in July 2026, is an open-weight multimodal model that powers video generation, Speech 2.8 produces lifelike voices, and Music 3.0 is an open-weight music model. People use these through MiniMax Code, MiniMax Design, a general agent and Hailuo AI, while developers use the API platform or the M Plan subscription. These Multimodal AI model families give one company's models for most media types, so it is worth comparing them as a set.
Open source: Yes. MiniMax publishes open weights on Hugging Face, including MiniMax M3, M3.1 Flash and Music 3.0, and describes H3 as an open model. (Checked on Hugging Face and official sites, November 2026.)
Who Should Use MiniMax?
- Developers: Use M3.1 Flash and M3 through the API, or MiniMax Code for coding.
- Video creators: Make videos with the H3 model in the Hailuo AI app.
- Audio creators: Generate voices with Speech 2.8 and music with Music 3.0.
- Researchers: Download open-weight models such as H3 and Music 3.0.
Professional reality: MiniMax is a Chinese company, and its consumer apps are run by different entities, such as Nanonoble Pte. Ltd in Singapore for Hailuo AI. Check the privacy policy of the specific product you use, and keep confidential data out of prompts.
MiniMax Features That Drive Business Results
MiniMax M3.1 Flash and M3
M3.1 Flash is built for professional agent work at speed and low cost, and M3 is the larger language model.
Result: capable agents and coding at a lower cost.
MiniMax H3 Open Model
H3 is an open-weight multimodal model that breaks the boundaries between tasks and modalities, and it powers Hailuo AI's video generation.
Result: high-quality video, with weights you can download.
MiniMax Speech 2.8
Speech 2.8 generates lifelike voices for narration, assistants and apps.
Result: natural voice-overs without a recording studio.
MiniMax Music 3.0
Music 3.0 is an open-weight, production-ready music model released in August 2026.
Result: original music for videos and apps.
MiniMax Code, Design and Agent
MiniMax Code (now with a Mini App), MiniMax Design and a general agent turn the models into ready-made tools.
Result: use the models without writing code.
API Platform and M Plan
Developers can call MiniMax models through the API platform or subscribe to the M Plan.
Result: one provider for text, video, voice and music APIs.
MiniMax Pricing in 2026
MiniMax sells API access on its platform and offers an M Plan subscription; current prices are shown there rather than on the homepage. Hailuo AI, its video app, gives new users free credits and has plans from $14.99 per month. Open-weight models such as H3 and Music 3.0 are free to download.
| Plan | Price | What You Get |
|---|---|---|
| API platform | From $0.15 per 1M input tokens | Access to M3.1 Flash, M3, H3, Speech 2.8 and Music 3.0 |
| Hailuo AI (video app) Best Value | From $14.99/month | Free credits for new users; video with MiniMax H3 |
| Open-weight models | Free to download | H3 and Music 3.0 weights |
If you are comparing Chinese AI providers by cost, you can read our DeepSeek review and our Qwen review for a closer look at how their pricing works.
Where MiniMax Is Strong / Where It Needs Care
- Models for every mediumText, video, speech and music models from one company.
- Open weightsH3 and Music 3.0 can be downloaded and run yourself.
- Agent-focused LLMM3.1 Flash targets fast, low-cost agent work.
- Large user baseMiniMax reports 300M+ individual users and 2M+ business clients and developers.
- Prices not on the homepageCheck the API platform for current model prices.
- Many separate productsMiniMax Code, Design, Agent and Hailuo AI each have their own sites and plans.
- Privacy varies by productDifferent apps are run by different entities; read each privacy policy.
- Professional RealityIf you only need one tool, such as video, a single-purpose app may be simpler.
Real-World Use Cases
Video Creation
Make short videos with the H3 model through Hailuo AI Video MiniMax's video app.
Voice-Overs
Generate narration with Speech 2.8 for videos, courses and apps.
Coding and Agents
Use M3.1 Flash or MiniMax Code to build and automate tasks at low cost.
Music for Content
Create background tracks with Music 3.0, or run the open weights yourself.
How to Get Started With MiniMax
Go to minimax.io to see the current models and products.
Pick a product: MiniMax Code, Design, the agent, or Hailuo AI for video.
Developers: sign up on the API platform and create a key.
To self-host, download open-weight models such as H3 or Music 3.0.
Is MiniMax Worth It in 2026?
MiniMax is worth following if you want strong Chinese models across text, video, speech and music from one company, with open weights for H3 and Music 3.0. Developers get an API and the M Plan, and creators can use Hailuo AI for video. It is less of a single product than a family of tools, so pick the one that fits your task and check its own pricing and privacy terms.
MiniMax vs the Competition
| Decision Area | MiniMax | When Another Option Wins |
|---|---|---|
| Best for | Models for text, video, speech and music | Kling AI if you want 4K options in a video app |
| Pricing | API and M Plan; Hailuo AI from $14.99/mo | DeepSeek for clearly listed per-token API prices |
| Key feature | Open-weight H3 and Music 3.0 | Qwen for a broader open-model range |
| Ease of use | Separate apps for code, design, agent and video | A single app if you only need one task |
| Scaling | API platform for developers | Self-hosting open models for full control |
MiniMax vs Runway
Runway is a US AI video suite with generation and editing tools. MiniMax offers video through Hailuo AI plus text, speech and music models.
Choose MiniMax if: You want models for several media types, including open weights. Choose Runway if: You mainly need video editing and effects in one app.
MiniMax vs DeepSeek
DeepSeek focuses on text models with a free chat and very low API prices. MiniMax covers more media, including video, voice and music.
Choose MiniMax if: Your workflow demands text, image, audio, and video from one dashboard. Choose DeepSeek if: You mainly need cheap text reasoning with a 1M-token context.
Frequently Asked Questions
Is MiniMax free?
Some parts are. Open-weight models like H3 and Music 3.0 are free to download, and Hailuo AI gives new users free credits. API use and the M Plan are paid.
What is MiniMax best used for?
Agents and coding with M3.1 Flash, video with H3, voices with Speech 2.8 and music with Music 3.0.
What is MiniMax H3?
MiniMax H3 is an open-weight multimodal model released in July 2026 that powers video generation, including in Hailuo AI.
Who runs Hailuo AI?
Hailuo AI is MiniMax's video app; its privacy policy names Nanonoble Pte. Ltd in Singapore as the operator.
What are the main limitations of MiniMax?
Prices are not on the homepage, products are split across several apps, and privacy terms differ by product.
Key Takeaways
- MiniMax builds M3.1 Flash, M3, H3, Speech 2.8 and Music 3.0.
- API and M Plan are paid; H3 and Music 3.0 weights are free to download.
- Best for developers and creators who want one company's models across media.
Best MiniMax Alternatives
- DeepSeek - DeepSeek is a Chinese AI chat assistant and low-cost API offering a 1M-token context window.
- Qwen - Qwen is Alibaba's model family offering open-weight models you can download plus an API on the Alibaba Cloud platform.
- StepFun - StepFun is a Chinese AI company offering Flash models, StepAudio voice models, open weights and an API platform.
- Claude - Anthropic's assistant Claude is strong at writing, analysis and coding, and offers an API for developers.
Bottom Line: MiniMax is a strong Chinese AI company with models for text, video, voice and music, several of them open-weight. Use Hailuo AI for video, MiniMax Code for coding, or the API for your own apps, and check each product's privacy terms.
Sources checked November 2026: MiniMax site, Hugging Face models.
Last Reviewed: November 2026 | Reviewed by theaitoolsbox.com editorial team
Pricing Plans
Free credits (Hailuo AI)
$0New users of the Hailuo AI video app receive free credits.
- Free credits for new Hailuo AI users
- Access to the Hailuo AI consumer video app
- No payment needed to start
Hailuo AI plans
From $14.99 per monthPaid plans for the Hailuo AI video app start at $14.99 per month.
- Plans start at $14.99 per month
- Consumer video app
- Free credits available for new users
M Plan
See minimax.io/m-planSubscription for individuals with three tiers: Go (1x usage quota), Explore (3x) and Build (7.5x). Prices are shown on minimax.io/m-plan, not in the docs.
- Three tiers: Go, Explore, Build
- Usage quotas of 1x, 3x and 7.5x
- M3.1 Flash text model on every tier
- H3 video model on Explore and Build only
- H3 Image and speech on all tiers
Free open weights
Free to downloadOpen-weight models such as H3 and Music 3.0 are free to download.
- H3 video model weights
- Music 3.0 weights
- Free to download
Prices checked November 22, 2026. Confirm on the vendor site before buying.
View Full Pricing on WebsiteAPI Pricing
MiniMax-M3.1-Flash
$0.15 input / $0.60 output per 1M tokensPay-as-you-go LLM pricing in USD. Prompt-cache reads cost $0.015 per 1M tokens.
MiniMax-M3
$0.30 input / $1.20 output per 1M tokensPermanent 50% off price for up to 512K input tokens; list price is $0.60 input and $2.40 output. Cache reads $0.06 per 1M tokens (list $0.12). Above 512K input tokens: $0.60 input, $2.40 output, $0.12 cache read (list $1.20 / $4.80 / $0.24). Priority tier is 1.5x the standard price.
H3 video (quality first)
$0.13 per second at 2K / $0.08 per second at 768PPriced per second of output video. The first 5 input images per task are free.
H3 video (Highspeed, Lite, Max)
$0.06 / $0.04 / $0.08 per second at 768PHighspeed is $0.06 at 768P and $0.04 at 480P; Lite is $0.04 at 768P and $0.025 at 480P; Max is $0.08 at 768P and $0.05 at 480P. First 5 input images per task are free.
Speech
$100 per 1M characters (hd) / $60 per 1M characters (turbo)speech-2.8-hd costs $100 per 1M characters and speech-2.8-turbo costs $60. Voice design is $3 per voice and rapid voice cloning is $1.5 per voice. Speech-to-text is $0.38 per hour.
API rates checked November 22, 2026. Rates can change; confirm on the vendor site.