← All companies

Frontier / Company profile

MiniMax

What MiniMax does

MiniMax (listed as MiniMax Group Inc.) is a multimodal foundation-model company that develops model families across language, vision/video generation, speech, and music, then distributes those capabilities through AI-native consumer products and an enterprise/developer open platform. The company positions its flagship model suite as multimodal and “agentic,” emphasizing ultra-long context capability and unified handling of text, audio, images, video, and music.

On the model side, MiniMax’s recent releases include MiniMax M3 (a coding/agentic model with up to a 1M-token context window using its MSA sparse-attention architecture), MiniMax H3 (general-purpose multimodal video generation, later open-sourced), and MiniMax Music 3.0 (an open-weights music model that generates an entire song in a single generation).

On the product side, MiniMax highlights a portfolio of AI-native applications such as Talkie (and Xingye), Hailuo AI (video generation), MiniMax Audio, an “Agent” desktop/product experience, and an open platform (API, token plan, and enterprise documentation).

Commercially, MiniMax reports user-scale traction via MAU (monthly active user) metrics across its AI-native applications and tracks Open Platform activity through average monthly active customers who make API calls.

News

Company record

Aug 13, 2026 · Official · MiniMaxMiniMax Music 3.0

MiniMax introduced MiniMax Music 3.0, describing it as an open-weights next-generation music generation model that generates a complete song (up to five minutes) from a creative concept and optional lyrics.

Aug 03, 2026 · Official · MiniMaxOpen General Intelligence: MiniMax H3 Is Now Open Source

MiniMax announced it is open-sourcing MiniMax H3 and described H3’s multimodal understanding and video generation specifications, including native stereo audio and supported output ranges, plus distribution details for released checkpoints and recommended inference tooling.

Jul 31, 2026 · Official · MiniMaxMiniMax H3: An Open Model Breaking the Boundaries Between Tasks and Modalities

MiniMax launched MiniMax H3 as a general-purpose multimodal generation model and described intended commercial content-creation use cases and technical components used in H3’s design.

Jun 29, 2026 · Official · MiniMaxMiniMax x Shanghai International Film Festival

MiniMax’s IR “New Releases” feed includes a dated entry for a MiniMax x Shanghai International Film Festival collaboration/news item.

Jun 09, 2026 · Official · MiniMaxMaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Evolutionary Search

MiniMax detailed its MaxProof framework for advancing mathematical proof capabilities, including improving a base model, aligning a verifier, building refinement capability, and designing a test-time scaling approach; it references prior benchmark results on IMO 2025 and USAMO 2026.

Jun 01, 2026 · Official · MiniMaxMiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model

MiniMax released MiniMax M3, describing frontier-level coding/agentic performance, MSA sparse attention supporting up to 1M tokens, and native multimodal capability for image/video input, plus links to experience routes via MiniMax Code/API services.

May 27, 2026 · Official · MiniMaxMiniMax Agent Team: Built for Long-Running Tasks and Continuous Evolution (renamed to Mavis)

MiniMax announced an upgrade to its Agent product, described new “Agent Teams” functionality (running multiple agents in parallel and coordinating roles), and explained a subscription unification between TokenPlan and Agent Plan with shared credits.

Apr 15, 2026 · Official · MiniMaxMiniMax at HumanX 2026

MiniMax reported participation at HumanX 2026 in San Francisco, including a featured talk by its President of Global Business, Linda Sheng, and descriptions of booth demos and stated user/enterprise scale claims.

Source map · 3 recurring channels · 15 references

Still resolving: Newsroom

Funding

Latest disclosed valuation$1.9K
Tracked capital$1.2B9 sourced rounds
DateRoundRaisedValuationLead / investorsEvidence
Jul 30, 2025Growth/extension round (amount disclosed)$300M
Mar 15, 2024Series Pre-B (Pre-IPO)$654$1.9K
Not attributed
Jun 01, 2023Series A (amount and valuation partly undisclosed)$250M$1.2B
May 04, 2023Series A (Pre-IPO)$257$900
Not attributed
Mar 29, 2022Pre-A Series (Pre-IPO)$50$500
Not attributed
Dec 02, 2021Angel Series (Pre-IPO)$31$169
Not attributed

What it builds

MiniMax M3

MiniMax M3 is a frontier coding and agentic model that uses MiniMax Sparse Attention (MSA) to support ultra-long context windows up to 1M tokens and provides natively multimodal support for image/video inputs.

source ↗
MiniMax H3

MiniMax H3 is a general-purpose multimodal generation model designed to understand unified multimodal context (text, images, video, and audio) and generate video with native stereo audio up to 2K resolution and up to 15 seconds.

source ↗
MiniMax Music 3.0

MiniMax Music 3.0 is an open-weights music generation model that composes, arranges, performs, and produces a complete song in a single generation, using a generation pipeline intended to sustain creative intent across a full song.

source ↗
MiniMax Agent (Mavis)

MiniMax Agent is an agent experience intended for long-running tasks; MiniMax announced an upgrade naming the upgraded agent “Mavis” and described the introduction of agent teams and a unified subscription across product surfaces.

source ↗
Talkie

Talkie is one of MiniMax’s AI-native multi-modal entertainment/creator products; MiniMax reports it as part of the Talkie/Xingye MAU segment.

source ↗
Xingye

Xingye is MiniMax’s AI-native multi-modal entertainment/creator product presented together with Talkie in MiniMax’s user metrics (Talkie/Xingye).

source ↗
Hailuo AI (video generation product)

Hailuo AI is MiniMax’s AI-native video generation product (referred to as “海螺视频” / Hailuo video in MiniMax’s IR site) and is included as a separate MAU segment in MiniMax’s reporting.

source ↗
MiniMax Audio (speech/voice/music experience)

MiniMax Audio is an AI-native audio product (referred to as “语音” and “MiniMax Audio” in MiniMax’s materials) and is tracked as its own MAU segment in MiniMax’s reporting.

source ↗
MiniMax open platform (API / enterprise developer platform)

MiniMax provides an open platform for enterprise and developers, including documentation, a token plan, and pricing; MAU and customer metrics in MiniMax’s disclosures include Open Platform activity measured by users making API calls.

source ↗

Key metrics

Milestones & partnerships