Practical cryptanalysis with MiniMax's latest Internal ModelOur research team used an internal-only research model that we will call MiniMax's Internal Model (MIM) to conduct automated cryptanalysis across modern cryptographic schemes. The model identified several previously unknown structural flawsMiniMax11/20 14:41
MiniMax H3-Image: Bringing 4K Image Generation into the 10-Second EraToday, we're releasing MiniMax H3-Image, an open-weight image generation model built on H3. It supports text-to-image generation, reference-based generation, and image editing. H3-Image can generate a 4K (4096 × 4096) image in under 10 secoMiniMax11/11 00:23
Introducing MiniMax Code Mini AppA Mini App is a long-running application with a frontend and a backend. It installs as a plugin, calls and is called by the Agent, and keeps the Agent’s work on a page that stays live. Today, we’re launching Mini App in the MiniMax Code desMiniMax10/22 14:32
MiniMax M3.1 Flash: Redefining Agent Performance and CostToday, we officially roll out MiniMax M3.1 Flash . Space Bunny 🐰, which processed more than 200 trillion tokens while operating anonymously, was an early version of MiniMax M3.1 Flash. As a next-generation Flash model, M3.1 Flash delivers MiniMax10/19 08:00
MiniMax|Japanese IP, Ready for the Global AI EraOn September 11, 2026, MiniMax and KAGAMIAI co-hosted the Japan IP.AI Co-Creation Conference in Shibuya, Tokyo. More than 150 companies from Japan and the United States joined us to explore the future of Japanese IP in the age of generativeMiniMax10/13 00:00
MiniMax Announces First Half 2026 Financial ResultsHONG KONG, Aug. 26, 2026 /PRNewswire/ -- MiniMax Group Inc. ("MiniMax" or the "Company"; HKEX: 00100), a leading global artificial intelligence company, today announced its unaudited financial results for the six months ended June 30, 2026.MiniMax08/26 19:36
MiniMax Music 3.0: Next-Generation Open-Weights, Production-Ready & Versatile Music ModelToday, we are introducing MiniMax Music 3.0, our next-generation music generation model. Given a creative concept and optional lyrics, the model composes, arranges, performs, and produces a complete song in a single generation. Music 3.0 foMiniMax08/13 21:33
Open General Intelligence: MiniMax H3 Is Now Open SourceToday, we are officially open-sourcing MiniMax H3, our next-generation general-purpose video model. System Overview MiniMax H3 is a general-purpose, omni-modal generative system. It supports unified understanding of multimodal contexts compMiniMax08/03 19:24
MiniMax H3: An Open Model Breaking the Boundaries Between Tasks and ModalitiesToday, we're launching MiniMax H3, a general-purpose multimodal generation model. H3 understands unified context across text, images, video, and audio, generating video with native stereo sound, up to 15 seconds at 2K resolution. Early testMiniMax07/31 08:00
MaxProof: Scaling Mathematical Proof with Generative-Verifier RL and Evolutionary SearchIntroduction In the M3 release post, we reported the performance of the M3 model on two international mathematical olympiad benchmarks: IMO 2025 and USAMO 2026. With the MaxProof framework, M3 exceeded the human gold-medal threshold on bothMiniMax06/09 21:43
MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One ModelMiniMax M3 is officially released today. M3 reaches frontier-level performance on specialized tasks such as coding and agentic work. It uses MSA (MiniMax Sparse Attention), a new attention architecture proposed by our team, and supports ultMiniMax06/01 01:31
MiniMax Agent Team: Built for Long-Running Tasks and Continuous EvolutionToday we are introducing the overall upgrade of MiniMax Agent. We have given the upgraded Agent a new name: Mavis — MiniMax as a Jarvis, your AI butler. This release brings the following updates: Launching Agent Teams. On MiniMax Agent deskMiniMax05/27 08:00
Why Can't the MiniMax LLM Say "Ma Jiaqi"? Internal Investigation of Sparse Token ForgettingBackground The MiniMax M2 series has attracted widespread attention from the developer community. During extensive usage, many users discovered certain corner cases — among which "the model cannot say Ma Jiaqi (马嘉祺)" sparked considerable diMiniMax05/26 00:30
MiniMax at the WaytoAGI Global AI ConferenceOn April 8th, we joined WaytoAGI at Happo-en in Tokyo for a global AI conference that brought together founders, builders, and operators from across the Chinese and global AI ecosystems. It was a chance to share how we think about multimodaMiniMax05/22 13:57
Imagination is ProductivityAs a global partner of World AI Film Festival (WAIFF 2026) and the co-initiator of the Chinese Special Track, MiniMax Hailuo took the center stage in Cannes. All greatness in human history begins with creation. We unveiled our 2026 brand viMiniMax04/25 00:00
MiniMax at Harvard XR 2026On April 11th, MiniMax joined over 20 speakers and hundreds of attendees at Harvard University for HXR 2026 , the annual conference organized by the Harvard GSD XR Club. This year's theme, "XR+: From Pixel to Voxel," centered on a question MiniMax04/16 17:31
MiniMax at HumanX 2026This past week, we joined thousands of builders, researchers, and operators at HumanX 2026 in San Francisco. It was our chance to share what we're building, hear what the industry is wrestling with, and connect with the teams shaping what cMiniMax04/15 11:29
MiniMax x Agora in Anime JapanDuring Anime Japan Week, MiniMax and Agora co-hosted an exclusive, business-focused summit in Tokyo, gathering top industry leaders, IP creators, and developers. The core mission is accelerating the evolution of AI characters from conceptuaMiniMax04/09 21:46
MiniMax AI Founder Day at GTCDuring NVIDIA GTC week, when the AI world converged in San Francisco, we hosted MiniMax AI Founder Day — bringing together a room full of founders, engineers, and builders shaping the next wave of AI. Fresh off the launch of M2.7, our firstMiniMax03/26 12:03
MiniMax M2.7: Early Echoes of Self-EvolutionIn the months following the first release of our M2-series models, we received a large volume of feedback and suggestions from enthusiastic users and developers, which drove us to further accelerate the efficiency of our model iterations. WMiniMax03/18 08:00
Forge: Scalable Agent RL Framework and AlgorithmForge RL framework hero Scaling reinforcement learning for real-world agents runs into a three-way conflict: system throughput , training stability , and agent flexibility all pull in different directions, and that tension has long blocked MiniMax02/14 20:00
MiniMax M2.5: Built for Real-World Productivity.Access API → Coding Plan → Try Agent Now → Today we're introducing our latest model, MiniMax-M2.5. Extensively trained with reinforcement learning in hundreds of thousands of complex real-world environments, M2.5 is SOTA in coding, agentic MiniMax02/12 08:00
A Deep Dive into the MiniMax-M2-herWorlds to Dream, Stories to Live How we built a Role-Play Agent for the production usage Three Years of Observations: How We Define Role-Play This year marks our third year optimizing Role-Play in Talkie / Xingye. Three years is long enoughMiniMax01/27 08:00
MiniMax M2.1: Significantly Enhanced Multi-Language Programming, Built for Real-World Complex TasksMiniMax has been continuously transforming itself in a more AI-native way. The core driving forces of this process are models, Agent scaffolding, and organization. Today we are releasing updates to the model component, namely MiniMax M2.1, MiniMax12/23 08:00
MiniMax M2 & Agent: Ingenious in SimplicityFrom Day 1 of our founding, we have been committed to the vision of "Intelligence with Everyone." Today, we are officially open-sourcing and launching MiniMax M2, a model born for Agents and code. At only 8% of the price of Claude Sonnet anMiniMax10/27 08:00