Homepage update: September 30, 2026 · MiniMax released M3.1-Flash-Preview on September 27, a 1M-context coding model with five effort levels, available only through its subscription plan and MiniMax Code for now.
MiniMax AI Chat: Independent Demo, Models, and API Guide
MiniMax is a global AI foundation-model company founded in early 2022. It develops models and products for language, coding, agent workflows, video, speech, image generation, and music.
MiniMax-AI.chat is an independent guide with a limited short-text demo. Use this page to test a focused prompt, understand the MiniMax AI ecosystem, and find the appropriate model, official product, API documentation, pricing guide, or safety resource for your goal.
Latest verified update (September 24, 2026): MiniMax now offers two current video models. MiniMax H3 is the general-purpose video-and-audio model, with text, image, video, and audio context and 4–15-second output at 768P or 2K; its H3-Base weights cover local 768P generation, not the complete open 2K workflow. MiniMax H3 Max, post-trained by fal.ai on H3 and released jointly by MiniMax and fal.ai, trades the 2K option for speed, with 5–15-second clips at 480P or 768P through the Pay-as-you-go API. Each guide documents access routes, pricing, and limits.
Try the MiniMax AI Chat Demo
This independent MiniMax AI chat demo is designed for brief text requests. It connects to one model configured by the site administrator; it does not reproduce every capability available through MiniMax products or the developer platform.
1,000 Characters
Each message can contain up to 1,000 characters. Keep one request focused and provide only essential context.
Text Only
The demo accepts typed prompts. It does not accept documents, images, audio, or video uploads.
One Configured Model
There is no visitor-facing model selector. Do not assume a response uses M3, M2.7, or another version unless the interface identifies it.
Independent MiniMax AI Assistant
Online
Hello! Ask me anything about MiniMax models, coding, or anything else.
Integration disclosure: Prompts are relayed through this website’s server to the API endpoint configured by the site administrator. The widget identifies the service as the MiniMax API, but its public interface does not expose the upstream route or configured model ID. For conversational continuity, recent turns may be sent with a later message. This site does not promise zero retention, private processing, or suitability for confidential data.
The demo works best for rewriting a short paragraph, summarizing a small excerpt, explaining a concept, improving a prompt, creating a brief outline, or discussing a compact code sample. It is not designed for long documents, model benchmarking, production evaluation, or high-stakes decisions.
Example prompt: “Explain this JavaScript error to a beginner. Give the likely cause, a corrected five-line example, and two checks I should perform next: [short error message].”
Privacy reminder: Do not submit passwords, API keys, financial details, health records, private customer information, confidential documents, or proprietary code. Read the Privacy Policy, Security information, and what not to paste into MiniMax AI before using the demo.
What Is MiniMax AI?
MiniMax describes itself as a global AI foundation-model company. Its portfolio spans general-purpose language models, coding and agent software, video generation, text-to-speech and voice tools, image generation, music generation, and developer APIs. The company’s products and its Open Platform serve different needs: an individual may want an official app or creative tool, while a developer may want an API key, model endpoint, usage quota, and commercial terms.
MiniMax-AI.chat serves a different role. It publishes independent explanations and provides a constrained public sandbox. Official MiniMax channels remain the correct destination for accounts, downloads, API keys, billing, product support, contracts, and binding policy information.
The current official product lineup includes MiniMax Code for coding-agent workflows, MiniMax Design and Hailuo Video for creative video work, MiniMax Audio for speech, voice, and music tools, and Talkie for character-based content and interaction. MiniMax also provides its Open Platform for developers. Its two current video models, MiniMax H3 and the faster MiniMax H3 Max, are available through supported official interfaces and the Video Generation V2 API. These names refer to different products and access paths, not one universal MiniMax chat interface.
The word “minimax” can also refer to the classic decision-making algorithm used in game theory and computer science. On this page, MiniMax means the AI company, its models, developer platform, and products—not the minimax algorithm.
MiniMax Models and Products at a Glance
MiniMax is broader than one chatbot. The table below is a dated map of the ecosystem, based on the official model catalog. These capabilities do not all exist inside the short-text demo above.
| Model or product | Primary role | Verified snapshot | Independent guide |
|---|---|---|---|
| MiniMax M3 | Language, coding, and agent workflows | Released June 1, 2026. The hosted API supports up to a 1,000,000-token combined context and native text, image, and video input. M3 returns text; specialist models handle media generation. | Read the M3 guide |
| MiniMax M3.1-Flash-Preview | Preview coding and agent model | Released September 27, 2026. 1,000,000-token context, text, image, and video input, and five effort levels (low to max, default max). Available only through MiniMax’s subscription plan (Token Plan, now M Plan) and MiniMax Code; no pay-as-you-go price and no official benchmarks yet. | Read the M3.1-Flash-Preview guide |
| MiniMax M2.7 | Engineering, office delivery, tools, and agents | Released March 18, 2026. The hosted model and high-speed variant remain listed with a 204,800-token context. M2.5 appears under legacy models in the official catalog. | Read the M2.7 guide |
| MiniMax H3 | Current video and native-audio generation, up to 2K | Released July 31, 2026. Hosted H3 accepts text, images, video, and audio context and generates 4–15-second clips at 768P or 2K with native stereo sound. H3-Base supports local 768P generation under a restricted community license; it is not the complete open 2K workflow. | Read the H3 guide |
| MiniMax H3 Max | Current high-speed video generation | Post-trained by fal.ai on MiniMax H3 and jointly released by MiniMax and fal.ai in late August 2026. Generates 5–15-second clips at 480P or 768P (no 2K) from text, first/last-frame images, or references, at $0.05 or $0.08 per output second through the Pay-as-you-go V2 API. | Read the H3 Max guide |
| Hailuo 2.3 and 2.3 Fast | Previous-generation video models | The current official API pricing catalog lists both models under legacy video models. Hailuo 2.3 supports text-to-video and image-to-video; 2.3 Fast supports image-to-video. They remain useful for historical comparisons and lower-cost workflows where supported. | Hailuo 2.3 and 2.3 Fast |
| Speech 2.8 HD and Turbo | Text-to-speech and voice workflows | Both variants are documented for 40 languages and seven emotions. The T2A API documents interjection tags for both speech-2.8-hd and speech-2.8-turbo; MiniMax describes HD as the ultra-realistic-quality option and Turbo as combining speed with natural flow. Check language, voice, emotion, and endpoint support together. | Read the Speech 2.8 guide |
| Music 3.0 | Song and instrumental generation | Released July 16, 2026. Effective August 20, 2026, MiniMax’s official service notice says paid Music Generation and Lyrics Generation APIs are no longer available to new users; existing paying users may continue, and the free Music APIs—including music-3.0-free—are discontinued. Eligible existing paying API users may continue with music-3.0 at POST https://api.minimax.io/v1/music_generation. New users should use MiniMax Audio or evaluate the separate MiniMaxAI/MiniMax-Music3 checkpoint for self-managed inference under the MiniMax-Music3 Community License. Music 2.6 remains a previous-generation model, but its free API variant is also discontinued. | Read the Music 3.0 guide |
| Image-01 | Image generation | Supports text-to-image and subject-reference workflows through hosted MiniMax access. Output dimensions, quantity, and temporary URL handling are governed by the image API. | Read the Image-01 guide |
| MiniMax Code and Agent Teams | Coding and multi-step task workflows | MiniMax Code is the official coding-agent product. Agent Teams can divide long tasks among agents. Feature availability can change by version, platform, account, and plan. | MiniMax Code guide and Agent guide |
For a status-based directory of language, video, speech, image, and music options, visit the MiniMax models hub. The hub separates recommended, available, previous-generation, and legacy entries so that older model searches do not distort guidance for new projects.
Recent independent tests: Review our 180-test MiniMax M3 Vision benchmark, 300-run tool-calling benchmark, and 300-image Image-01 text benchmark. We have not yet completed an independent H3 or H3 Max benchmark; both video guides label provider examples and speed claims as vendor evidence.
Choose the Right MiniMax Access Path
The best access path depends on whether you want a quick experiment, an official end-user product, an application integration, or downloadable model weights. Do not treat these routes as interchangeable.
| Access path | Best for | Important limitation |
|---|---|---|
| MiniMax-AI.chat demo | Brief, low-risk text exploration without a site account | Independent interface; 1,000 characters; one configured model; no uploads or model selector |
| Official MiniMax products | Coding, agent tasks, video creation, audio tools, and character experiences | Features depend on product, release, platform, geography, account, and plan |
| MiniMax Open Platform | Developers who need API keys, model selection, quotas, logs, and production integration | Requires registration and acceptance of official pricing, terms, and technical limits |
| Open-weight models | Teams evaluating downloadable M-series weights, H3-Base, MiniMax-Music3, and self-managed deployment | “Open-weight” does not mean unrestricted open source. MiniMax-Music3 uses its own Community License, and its checkpoint ID and local serving route are not interchangeable with the hosted Music API IDs and endpoint. H3-Base also does not include the complete 2K workflow. |
For video, creators can use official interfaces such as MiniMax Design, which offers both H3 and H3 Max, or Hailuo Video for H3. Developers can call either model through the Open Platform Video Generation V2 API with Pay-as-you-go billing, and eligible teams can evaluate H3-Base weights for local 768P deployment. Third-party hosts such as fal also serve H3 Max under their own prices and terms. Hosted 2K H3, the faster H3 Max, and local H3-Base are not equivalent, so review the H3 access and license guide and the H3 Max guide before choosing a route.
Use our MiniMax app and official-access guide to find end-user destinations. Use the API guide for development, and read the relevant model license before any self-hosted or commercial deployment.
What Can You Do With MiniMax AI?
The wider MiniMax ecosystem covers several kinds of work. Availability inside an official product or API endpoint depends on the selected model and plan; only brief text requests are available through the independent demo on this page.
- Write, reason, and code: use an M-series language model for explanations, drafting, software tasks, tool calls, structured output, and supported multimodal analysis.
- Build agent workflows: use coding or agent products for multi-step tasks, with product-specific tools, permissions, and availability limits.
- Generate video and native audio: use MiniMax H3 for multimodal video generation up to 2K, or MiniMax H3 Max for faster 480P and 768P clips. The H3 guide explains when to choose an official product interface, the V2 API, or locally deployed H3-Base weights.
- Create speech and voices: use Speech models for text-to-speech, supported voice cloning, voice design, streaming, or asynchronous long-text generation.
- Generate images: use Image-01 for supported text and subject-reference image workflows.
- Create music: existing paying API users may continue to use hosted Music 3.0, but MiniMax says paid Music Generation is closed to new users from August 20, 2026 and the free Music API variants are discontinued. New users should use MiniMax Audio or evaluate
MiniMaxAI/MiniMax-Music3for self-managed local inference. The hosted, product, and local routes use different identifiers, endpoints, request schemas, output behavior, infrastructure, and legal terms.
Explore practical examples in the MiniMax AI use-cases hub, including code generation, business workflows, YouTube production, customer support, education, and marketing agencies. Each guide should be treated as a starting point for testing, not a guarantee of fitness for a regulated or high-stakes workflow.
MiniMax API and Pricing
The official MiniMax API overview covers language, speech, video, image, music, and file-management capabilities. Developers register through the Open Platform and create the appropriate credential. Pay-as-you-go access uses an API key, while Token Plan subscriptions use a separate subscription key.
For text models, MiniMax documents Anthropic-compatible and OpenAI-compatible routes. Media generation uses modality-specific APIs, so one chat request does not automatically generate video, speech, images, and music. Start with the MiniMax API guide, then use the application-building guide and API error-code reference for implementation details.
MiniMax H3 and H3 Max use the separate Video Generation V2 API and are not available through the short-text demo on this homepage. Current rates, input charges, and operational limits are maintained in the H3 guide, the H3 Max guide, and the dated pricing guide.
MiniMax pricing varies by model, input and output usage, media settings, speed tier, subscription structure, and sometimes context length. Instead of copying a large rate table onto the homepage, compare the dated MiniMax pricing guide with the official pricing overview before spending money. If you are comparing monthly quotas with usage-based billing, read the MiniMax Token Plan guide.
Privacy, Safety, and Responsible Use
Separate three questions before using MiniMax AI: what this independent demo processes, what an official MiniMax product processes, and what your own application stores. The policies and data flows are not identical.
- Minimize submitted data. Remove personal, confidential, regulated, security-sensitive, and proprietary information before sending a prompt.
- Expect mistakes. AI output can be incomplete, inaccurate, outdated, insecure, or unsuitable for the intended audience.
- Verify high-impact claims. Do not rely on an AI response as medical, legal, financial, employment, compliance, or security advice.
- Review the relevant terms. Product, API, and model-weight licenses impose different responsibilities and restrictions.
- Test your own workload. Public benchmarks and vendor examples do not replace evaluation with representative data, failure cases, budgets, and human review.
MiniMax’s official API privacy policy and official API terms govern official developer-platform use; this website’s Privacy Policy governs this independent interface. We do not claim zero retention, guaranteed data residency, HIPAA compliance, SOC 2 certification, or universal GDPR compliance.
For a fuller decision framework, read Is MiniMax AI safe?, using MiniMax AI safely at work, and our MiniMax AI and GDPR guide.
Compare MiniMax AI With Other Platforms
No single platform is the best choice for every workload. Compare the exact model version, input types, context limits, tool support, latency, price, deployment route, privacy requirements, and failure modes that matter to you. Start with the MiniMax AI comparison hub or review focused comparisons with ChatGPT, DeepSeek, Claude, and Google Gemini.
MiniMax AI FAQ
What is MiniMax AI?
MiniMax is an AI foundation-model company founded in early 2022. It develops language and coding models, agent software, video tools, speech and voice systems, image generation, music generation, and a developer platform.
Is MiniMax AI Chinese?
Yes. MiniMax is a Chinese AI company headquartered in Shanghai, while its products and developer platform serve international users. The company describes itself as a global AI foundation-model company. See our sourced MiniMax company, country, and ownership guide. MiniMax-AI.chat is a separate independent website.
Is MiniMax-AI.chat the official MiniMax website?
No. MiniMax-AI.chat is an independent publication and limited demo. It is not affiliated with, endorsed by, sponsored by, or operated by MiniMax.
Which model powers this MiniMax AI chat demo?
The widget uses one model selected in the site’s administrative configuration, but the public interface does not expose its model ID. The page therefore does not label the demo as M3, M2.7, or another version without visible technical confirmation.
Can I switch MiniMax models in this demo?
No. The interface has no visitor-facing model selector. Use the model guides for comparison and an official MiniMax access path where the required model is supported.
Does this demo provide M3’s full 1M context window?
No. The public interface limits each message to 1,000 characters. MiniMax documents up to a 1,000,000-token combined context for hosted M3 access, but that capability is not exposed by this demo.
Can I upload images, documents, audio, or video?
Not here. This demo accepts text only. Supported official models and products offer other input or generation types through separate interfaces and endpoints.
Is MiniMax AI free?
The short-text demo on this page is free to try, but it is not an H3 or H3 Max video model and each message is limited to 1,000 characters. Official MiniMax products may offer trials or credits, while API usage and some plans are paid. Open-weight downloads also carry model-specific licenses and infrastructure costs. Check the current pricing guide before relying on free access.
How much does MiniMax AI cost?
There is no single MiniMax price. Language models are generally billed by tokens, while video, speech, image, and music use modality-specific rates. On September 24, 2026, MiniMax listed H3 output at $0.08 per second for 768P and $0.13 per second for 2K, and H3 Max output at $0.05 per second for 480P and $0.08 per second for 768P, before applicable input charges. Speech Recognition was listed at $0.38 per hour. Compare the MiniMax pricing guide with the H3 cost examples and H3 Max cost examples.
What is MiniMax H3?
MiniMax H3 is MiniMax’s general-purpose video-and-audio generation model, released July 31, 2026, and one of its two current video models alongside H3 Max. Hosted H3 supports 4–15-second output at 768P or 2K with native stereo audio and multimodal references. Public H3-Base weights support local 768P generation under a restricted license; the complete 2K workflow is not open. See the verified H3 guide.
What is MiniMax H3 Max?
MiniMax H3 Max (MiniMax-H3-Max) is a speed-optimized video model that fal.ai post-trained on MiniMax H3; MiniMax describes it as jointly released by MiniMax and fal.ai. It generates 5–15-second clips at 480P or 768P, without a 2K option, from text, first/last-frame images, or references, and it is billed through MiniMax’s Pay-as-you-go API at $0.05 or $0.08 per output second. See the H3 Max guide for pricing, access routes, and how it compares with H3.
What is MiniMax M3.1-Flash-Preview?
MiniMax M3.1-Flash-Preview (MiniMax-M3.1-Flash-Preview) is MiniMax’s newest M-series language model, released September 27, 2026 for coding, tool use, and agent work. It has a 1,000,000-token context window, accepts text, images, and video, and lets you set thinking depth with five effort levels, defaulting to max. For now it is available only through MiniMax’s subscription plan (Token Plan, now M Plan, from $22 a month) and MiniMax Code. It has no pay-as-you-go price, and MiniMax has not published benchmarks. See the M3.1-Flash-Preview guide.
Where can I get an official MiniMax API key?
Register through the official MiniMax Open Platform and create the credential required for your billing method. MiniMax-AI.chat does not issue API keys, sell official API access, manage billing, or provide account support.
Is MiniMax AI better than ChatGPT?
Neither is universally better. MiniMax is a model-and-product ecosystem, while ChatGPT is an OpenAI product. MiniMax H3 is designed specifically for multimodal video generation with native audio; the better choice for chat, coding, research, or agent workflows depends on the exact models, tools, privacy requirements, and price. Use our MiniMax vs ChatGPT comparison for a workload-based decision.
Can I use MiniMax models commercially?
It depends on the access route, model, territory, and organization. Hosted products and APIs follow their service terms. Downloadable weights use model-specific licenses. H3-Base has explicit territory, attribution, revenue, and model-use restrictions. The MiniMax-Music3 Community License requires commercial products or services using the weights to display “MiniMax-Music3” prominently, and it requires prior written MiniMax authorization if aggregate yearly revenue from those products or services, including affiliates, exceeds US$20 million. Do not treat public weights as unrestricted open source; review the exact current license and Acceptable Use Policy.
Can I use the demo for confidential or high-stakes work?
No. Do not submit confidential, personal, regulated, or security-sensitive information. Verify important output with reliable sources and qualified professionals.
Is MiniMax AI the same as the minimax algorithm?
No. MiniMax AI refers here to the company and its AI ecosystem. The similarly named algorithm is a separate method for choosing moves under adversarial assumptions.
How We Verify MiniMax Information
We prioritize operational sources over undated marketing summaries: the live API catalog and official model release notes, followed by API references, pricing documentation, model repositories and licenses, dated announcements, and general product pages. This order matters because MiniMax pages can be updated at different times.
This homepage’s video information was last updated on September 24, 2026, when MiniMax H3 Max was added and H3 specifications, access routes, pricing, and current-model status were rechecked against MiniMax’s video guide, V2 API documentation, pricing catalog, rate-limit page, and official website. Other model snapshots retain the dates stated in their dedicated guides. We distinguish MiniMax’s published claims from independently observed limitations and preserve older-version guides only where they serve a real search need.
Learn more about MiniMax-AI.chat, read our Disclaimer, or request a correction. For production decisions, verify model availability, API limits, pricing, licenses, privacy conditions, and commercial terms through official MiniMax channels.
Explore MiniMax AI
Try a brief prompt, browse the model directory, or continue to the API and pricing guides that match your goal.