MiniMax
MiniMax is a Chinese artificial intelligence technology company focused on the research and development of large language models and multimodal AI technologies. The company independently develops a series of foundational large language models and builds an AI capability platform for enterprises and developers, offering AI services including text generation, speech synthesis, and image understanding. MiniMax is a representative company among China's new generation of large model startups.
Core Facts
MiniMax H3 is a multimodal video generation model capable of simultaneously generating video frames and matching audio (including dialogue, sound effects, and ambient sound).
Prompt Studio can generate structured MiniMax H3 video scripts.
H3 Minimax is Minimax's video generation engine, which adopts a Sparse Attention mechanism.
Timeline (last 90 days)
MiniMax has open-sourced Mcode Stack, the workflow the MiniMax Code team uses daily, covering code review, sub-agent splitting, context budgeting and long-running tasks; the repository uses the MIT license and does not include the source code of MiniMax Code itself.
MiniMax open-sourced Mcode Stack, disclosing the agent workflows and skill collection its development team uses daily, including project review, context management and palm-task handling, but the repository does not contain the original code of MiniMax Code.
vLLM provided Day-0 support for MiniMax M3.1 Flash on the day it was released.
@inferact provides SLA-guaranteed inference hosting services for MiniMax M3.1 Flash.
Engineering improvements related to MiniMax M3.1 Flash are being upstreamed to vLLM.
Domestic open-source models such as Huawei, Kimi, MiniMax, and DeepSeek made a concentrated appearance at this year's Wuzhen conference.
For monthly entry-tier pricing, the order is Zhipu highest, Kimi in the middle, and MiniMax lowest; the middle tier and the highest tier follow the same order.
For MiniMax's M2.5 and M2.7 series, starting October 19 requests will be uniformly routed to M3.1 Flash Preview, and the series will be officially discontinued on November 18; the official statement says existing package usage will not be affected.
MiniMax has two refresh windows: a 5-hour window plus a 7-day weekly window, both starting from first use and refreshing upon expiry, with unused quota not accumulating; non-video models require remaining quota in both windows to be usable.
MiniMax's packages changed from Token Plan to M-Plan, with three tiers: Go, Explore, and Build; for quotas the official information only states multiples: Explore is 3 times Go, and Build is 12 times.
40 more timeline events
All Facts (20)
MiniMax H3 is a multimodal video generation model capable of simultaneously generating video frames and matching audio (including dialogue, sound effects, and ambient sound).
90%VerifiedPrompt Studio can generate structured MiniMax H3 video scripts.
80%VerifiedH3 Minimax is Minimax's video generation engine, which adopts a Sparse Attention mechanism.
80%VerifiedWith the accelerator enabled, MiniMax H3 takes approximately 2 minutes to generate 1 second of video, approximately 14 minutes for a 480P ten-second video, and approximately 20–30 minutes for a 768P ten-second video.
80%VerifiedMiniMax-H3 features two capabilities: image-to-video generation and reference-based generation.
80%VerifiedM3.1 Flash can be called via the AI SDK, Chat Completions, Responses, and the Anthropic Messages API, with the model name minimax/minimax-m3.1-flash.
75%VerifiedDeepSeek is deploying native multimodal capabilities at the Flash level rather than the flagship version first, with the intent to prioritize capturing high-frequency, large-scale application scenarios.
75%VerifiedMinimax is one of the leading large model companies in China, and its video generation product line includes "Hailuo."
75%VerifiedThe reference-image-to-video workflow supports reference image, video, and audio input, with lip sync and audio highly synchronized in the demo.
70%VerifiedThe official name of MiniMax M3.1 Flash is `MiniMax-M3.1-Flash`, positioned as a frontier multimodal coding model with a 1 million context window and adjustable thinking depth.
65%VerifiedMiniMax M3.1 Flash Preview emphasizes speed and can be used for quickly fixing bugs in code development.
65%VerifiedMiniMax's video generation capability is primarily based on a Diffusion Model architecture.
65%VerifiedThis workflow can generate continuous videos up to 60 seconds long on a 12 GB GPU.
65%VerifiedMiniMax has a presence across multiple AI modalities including large language models (the abab series), speech synthesis, and video generation
65%VerifiedMiniMax is a China-based AI startup focused on the field of large language models and multimodal models.
65%VerifiedMiniMax's video generation product Hailuo AI was launched in 2024.
65%VerifiedThe Wan video generation model supports two generation modes: Text-to-Video and Image-to-Video.
65%UnverifiedMiniMax的模型架构采用了混合专家（MoE）与线性注意力的创新组合
80%Unverified月之暗面（Kimi母公司）、智谱AI（GLM母公司）、MiniMax均有阿里巴巴的投资背景
80%Partially VerifiedMiniMax M3.1 Flash is available on AIHubMix for token-billed calls under the model ID `minimax-m3.1-flash`, with input priced at $0.165 per million tokens and output at $0.66 per million tokens.
65%