Create detailed AI images with MiniMax H3 Image. Generate from text, edit existing images, blend up to 16 references, and explore high-resolution 1K, 2K, and 4K outputs.
MiniMax H3-Image is an open-weight model for native 16MP image generation and editing. It uses just six diffusion steps, handles text-to-image and editing in one checkpoint, and supports up to 16 reference images.
MiniMax H3 is an open multimodal model that generates 2K video with native stereo sound. It unifies text, image, and audio inputs, excelling at accurate text rendering, visual packaging, and complex instruction following for commercial content creation.
MiniMax Design is a desktop app that wraps MiniMax H3 in a full production workflow. Describe the video, and agents handle the brief, stills, H3 clips, music, and edit on one canvas. Skills, a 3D director stage, and ComfyUI nodes stay there when you want more control.
MiniMax M2.7 is a self-evolving AI model that helped build its own capabilities. It can create agent harnesses, collaborate via Agent Teams, and handle complex tasks like coding, debugging, and research. With strong SWE-Pro performance and reduced intervention time, it moves beyond static AI into systems that continuously learn, adapt, and execute complex work with minimal human input. Available via API and MiniMax Agent for builders pushing AI-native workflows.
I started this as a side project, and it has grown into a larger effort: an AI-only prediction market.
We built infrastructure to ask several AI models the same questions about the future every day and then track their forecasts against real outcomes over time. Some of these questions are similar to markets on Polymarket, which let us test a simple hypothesis:
Are LLMs more rational than the prediction market crowd?
Launched last week, open-source frontier model @MiniMax M2.7 scores 56.2% on SWE Bench Pro, converging towards the best proprietary models like @Claude by Anthropic Opus 4.6.
Introducing M2.5, an open-source frontier model designed for real-world productivity. SOTA performance at coding (SWE-Bench Verified 80.2%), search (BrowseComp 76.3%), agentic tool-calling (BFCL 76.8%) & office work. Optimized for efficient execution, 37% faster at complex tasks. At $1 per hour with 100 tps, infinite scaling of long-horizon agents now economically possible.
MiniMax M3 reaches frontier-level performance on coding and agentic tasks, with a 1M context window powered by the MSA architecture and native multimodal understanding. The first open-weight model to unite all three capabilities.
