MiniMax AI Statistics By Features, Pros And Cons (2026)

Introduction

MiniMax AI Statistics: MiniMax AI is a China-based artificial intelligence company that is rapidly expanding its development on large language models and multimodal AI systems. It designs and delivers AI-powered tools, including chat assistants and content generation platforms, that individual users and enterprises worldwide widely use. The company is still in the early stages of generating revenue, but it is growing quickly in terms of revenue, annual recurring revenue, and user base. With strong global expansion and rising adoption, MiniMax AI is becoming an important new competitor in the generative AI space, alongside leading global AI labs.

Editor’s Choice

1. MiniMax reported approximately USD 79 million in 2025 revenue, with around 158.9% year-on-year growth from USD 30.5 million in 2024.
2. In 2026, revenue is expected to reach USD 183.9 million, gross profit USD 44.8 million, and loss USD 534.6 million.
3. The company’s momentum continued, with annual recurring revenue exceeding USD 150 million in February 2026.
4. MiniMax expanded its ecosystem to 236 million+ cumulative users by the end of 2025 across 200+ countries and regions.
5. Gaming AI is the largest use case for Minimax, accounting for 40% of total market demand.
6. Office productivity tools use advanced models like M2.7 to support complex Excel, Word, and PowerPoint tasks, document writing, and workflow automation more efficiently.
7. As of March 2026, MiniMax.io received around 7 million total visits, up by 12.7% from last month.
8. China accounts for the largest share of traffic, at about 12.18%, down 23.98% from last month.
9. The largest age group is 25-34, representing 35.97% of all visitors to the site.
10. The M2.7 model uses around 10B active parameters and shows strong benchmark results, including 56.22% on SWE-Pro, 78% on SWE-bench Verified, and 57% on Terminal Bench 2.
11. The M2.7 Standard model has a context window of up to 205K tokens, with input cost at USD 0.30 per 1M tokens, output cost at USD 1.20 per 1M tokens, and an approximate blended cost of USD 0.53 per 1M tokens at a 3:1 input-output mix.
12. In head‑to‑head tasks against Claude Sonnet 4.5, the M2 agent wins 27.9% of tasks, ties on 34.9% of tasks, and loses 37.2% of tasks.

MiniMax Company Overview

Key Features

• M2.5 and M2.7 are agentic models that can plan tasks, execute steps, and verify outputs, making them useful for coding and office workflows.
• The platform supports omni-modal input and output, including text, speech, images, video, and music, within a single model ecosystem.
• It can process very long contexts of up to 1 million tokens, enabling work on large documents, codebases, and long-horizon tasks.
• It uses a Mixture-of-Experts architecture and lightning attention to improve speed and reduce compute costs for large inputs.
• The models are strong in software engineering, supporting full-stack development, debugging, and production-level application building.
• Access is provided through APIs and cloud integrations such as Vertex AI, AWS, and GCP for easy enterprise deployment.

MiniMax AI Financial Report

• MiniMax reported approximately USD 79 million in 2025 revenue, with around 158.9% year-on-year growth from USD 30.5 million in 2024, according to Investing.com.
• Meanwhile, the stock surged up to approximately 21% intraday.
• The company’s momentum continued, with annual recurring revenue exceeding USD 150 million in February 2026, according to the South China Morning Post reporting.
• A report published by Yahoo Finance, with over 70% of revenue coming from international markets.
• On a normalized basis, adjusted net loss was USD 250.9 million.
• MiniMax raised approximately USD 618-620 million via its Hong Kong IPO
• The official website of MiniMax also shows profitability improved at the gross level, with gross profit of USD 20.1 million (+437%) and gross margin rising to 25.4% from 12.2% in the prior year.
• Despite operational improvement, the company reported a net loss of USD 1.87 billion (IFRS).

(Source: substackcdn.com)

• In 2026, revenue is expected to reach USD 183.9 million, gross profit USD 44.8 million, and loss USD 534.6 million.
• By 2027, revenue may grow to USD 356.2 million, gross profit to USD 97.5 million, and loss may narrow to USD 477.2 million.

MiniMax User Scale

• MiniMax expanded its ecosystem to 236 million+ cumulative users by the end of 2025 across 200+ countries and regions, according to its official website.
• Moreover, it also served about 214,000 enterprise customers and developers.
• By April 2026, the company will be scaling to over 250 million users and 200,000+ enterprise clients.
• Tech Buzz China Insider reported that monthly active users totaled 27.6 million, including 1.77 million paying users with an ARPPU of approximately USD 15.

Minimax Use Cases

• Content creators and YouTubers use Minimax AI to turn scripts into short videos with visuals and voiceovers for storytelling and explainers.
• Marketing teams create promotional videos from blog posts or product pages and test versions on TikTok, Instagram, and LinkedIn.
• Education, corporate, media, gaming, influencers, and startups use it for training videos, multilingual learning, news summaries, social content, and cost-efficient investor or branding videos.

By Industry Demand

(Source: promptaitools.com)

• Gaming AI is the largest use case for Minimax, accounting for 40% of total market demand.
• Decision-making AI is the second-largest segment at 25%, and robotics and path planned to contribute 15%.
• Cybersecurity and risk analysis accounted for 10% of threat detection, followed by stock market prediction, which accounted for 10% of forecasting opportunities.

Real-World Application Domains Overview

• Office productivity tools use advanced models like M2.7 to support complex Excel, Word, and PowerPoint tasks, document writing, and workflow automation more efficiently.
• According to a report published by Remote Openclaw Blog, software engineering systems achieve 56.2% on SWE-Pro and 57.0% on Terminal Bench 2, enabling code generation, debugging, and full project delivery, with VIBE-Pro at 55.6%.
• AI agents such as MiniMax Agent support multi-step research, structured reporting, and multimodal content creation for market research and business automation tasks.
• Multimodal and XR applications shown at HumanX and Harvard XR 2026 demonstrate interactive entertainment, voice and vision-based workflows, and next-generation XR experiences.

MiniMax Website Traffic And Performance (March, 2026)

• According to SimilarWeb, MiniMax.io received around 7 million total visits, up by 12.7% from last month.
• The website held a global rank of #6,998 and ranked #3,264 in India.
• Moreover, the user engagement is healthy with a 34.69% bounce rate, 5.25 pages per visit, and an average session duration of 4 minutes 47 seconds.
• On minimax.io, 73.41% visits are from desktops and 26.59% from mobile devices globally.

By Country

(Reference: similarweb.com )

• China accounts for the largest share of traffic, at about 12.18%, down 23.98% from last month.
• The United States, Vietnam, Brazil, and Brazil also show meaningful adoption, with 11.69% (+3.77%), 6.64% (+25.75%), 6.3% (-13.09%), and 5.61% (-2.06%) of traffic, respectively.
• Meanwhile, the remaining 57.57% traffic share comes from other global regions.

By Demographics

(Reference: similarweb.com )

• Most visitors to minimax.io are men, making up about 66.81% of the audience, while women account for 33.19%.
• The largest age group is 25-34, representing 35.97% of all visitors to the site.
• Younger adults aged 18-24 make up 18.71% of visitors, while older groups aged 35-44, 45-54, 55-64, and 65+ make up 20.89%, 12.92%, 7.43%, and 4.08% of traffic, respectively.

MiniMax 2026 Models Portfolio Overview

• MiniMax’s 2026 AI lineup is led by M2.7, which is positioned as its flagship agentic model, designed for complex reasoning and advanced tasks.
• The M2.7 model uses around 10B active parameters and shows strong benchmark results, including 56.22% on SWE-Pro, 78% on SWE-bench Verified, and 57% on Terminal Bench 2, indicating competitive performance in coding and agent tasks.
• Its GDPval-AA ELO score of 1495 and 97% task adherence across enterprise workflows highlight strong reliability in structured work environments.
• The M2.5 model serves as the main general-purpose engine, achieving 80.2% on SWE-bench Verified, close to leading frontier models, while being faster and more cost-efficient.
• M2.5 is also reported to generate a significant share of internal production code and to operate at high speed with a low cost per unit of compute.
• The Lightning variant of M2.5 supports up to 1 million token context windows and is optimized for high-throughput applications and lower per-token pricing.
• Older models like M2.1 and M2 remain available, mainly for cost-sensitive or legacy compatibility use cases.

MiniMax AI Model Pricing Analysis

Agent Benchmark Results And Cost Efficiency Comparison

(Source: minimax.io)

• In head‑to‑head tasks against Claude Sonnet 4.5, the M2 agent wins 27.9% of tasks, ties on 34.9% of tasks, and loses 37.2% of tasks.
• Against GLM 4.6, the M2 agent wins 42.9% of tasks, ties on 31.0% of tasks, and loses 26.2% of tasks.
• Against Kimi‑K2‑0905, the M2 agent wins 64.3% of tasks, ties on 21.4% of tasks, and loses 14.3% of tasks.
• Against DeepSeek V3.2, the M2 agent wins 71.4% of tasks, ties on 19.0% of tasks, and loses 9.5% of tasks.
• The average cost per task is about USD 0.55 for M2, around USD 4.41 for Sonnet 4.5, roughly USD 1.10 for GLM 4.6, approximately USD 1.48 for Kimi‑K2‑0905, and about USD 0.19 for DeepSeek V3.2.

Full‑Stack Development Performance And Cost

Terminal Use Performance And Cost

Minimax AI Pros And Cons

• Minimax AI produces high-quality video generation with realistic details and supports multiple languages for global use. It has an easy-to-use interface for beginners and offers real-time visual and audio customization, including AI voiceovers.
• However, complex prompts may confuse the system, processing can be slow during high traffic, and it has less innovation than top tools. Some advanced features also need improvement.

MiniMax AI Scorecard Summary

• According to All About AI, the ease of use scores 4.6/5, with simple drag-and-drop and an easy workflow.
• Motion realism is 4.7/5 with natural gestures and expressions.
• Video generation speed is 4.2/5, with a fast 6-second output; creative control is 4.3/5; and model flexibility is 4.5/5 with multimodal input.
• Language support is 4.4/5, business features 4.1/5, and collaboration tools 4.0/5.

Conclusion

MiniMax AI is emerging as a fast-growing global artificial intelligence company with strong user adoption and expanding enterprise presence. It has built a wide international footprint through its consumer and business AI products. While the company is still investing heavily and not yet consistently profitable, its rapid growth in users, improving product ecosystem, and rising commercial adoption suggest strong long-term potential in the competitive AI foundation model space.

FAQ