Build Agents. Create Worlds. All for Free.
Today, we’re excited to announce GMI Cloud has been selected as a first-tier global launch partner for MiniMax, bringing day-0 access to both MiniMax M2 and MiniMax Hailuo 2.3 / 2.3-Fast—now available to deploy for free on the GMI Cloud Console.
This dual release bridges two frontiers of intelligence: agentic reasoning and generative imagination.
- M2 is a model built for code, tools, and autonomous agents.
- Hailuo 2.3 is the next-generation text-to-video model that delivers cinematic realism and human-level physics.
Together, they redefine what’s possible when builders and creators share the same computational canvas.
And we’re celebrating this milestone with a joint launch event on October 31st, 2025 in Mountain View, California — an in-house party with creators, researchers, and builders exploring the edge of agentic and cinematic AI.
🔗 Join us on Luma
MiniMax M2 — Born for Agents, Built for Speed
MiniMax M2 is an open-sourced intelligence engine designed for end-to-end agentic workflows.
From code generation to shell execution, from research automation to long-chain tool calls, M2 demonstrates performance that rivals frontier models at a fraction of the cost.
- 8% of Claude 3.5 Sonnet’s price, at 2× the inference speed
- Ranked Top-5 globally on the Artificial Analysis Index (AAI 61)
- Excels on SWE-Bench Verified (69.4), GAIA (75.7), and BrowseComp (44.0)
- Built for deep integration with developer workflows like Cursor, Claude Code, Cline, and Droid
Agent benchmarks show M2 outperforming peers such as DeepSeek V3.2 and GLM-4.6 in over 70 % of tasks, while maintaining cost-efficiency down to $0.55 per task.
For developers building copilots, research assistants, or multi-tool orchestrators, M2 strikes the ideal balance between intelligence, speed, and affordability. M2 can be deployed directly on GMI Cloud with either a dedicated endpoint or MaaS option.
Weights are open-sourced on Hugging Face for local deployments.
→ Deploy MiniMax M2 for free on GMI Cloud
MiniMax Hailuo 2.3 & 2.3-Fast — Physics, Motion, and Imagination Unbound
Where M2 builds the brains for agents, Hailuo 2.3 brings the vision.
The new model pairs advanced human-motion physics with cinematic camera control to generate scenes that look and feel alive.
Highlights:
- Supports both text-to-video and image-to-video generation
- Resolutions: 768 p / 1080 p, durations 6–10 seconds (6 s for 1080 p)
- Hailuo 2.3-Fast offers ~55 s latency for 768 p renders
- Enhanced VFX realism and style transformation (Pixar-style, surreal lighting, water-light effects)
- Improved face and object consistency, text & logo animation, and prompt adherence
Early creators describe Hailuo 2.3’s motion quality as “so realistic it could be shown on TV.” Its handheld camera movement, subtle micro-shakes, and emotional expressiveness mark a leap forward for AI-generated film and animation.
Whether you’re storyboarding a music video, prototyping a cinematic trailer, or animating anime-style sequences, Hailuo 2.3 delivers professional-grade realism accessible to anyone.
→ Generate videos with Hailuo 2.3 for free on GMI Cloud
MiniMax Speech 2.6 — Ultra-Fast, Ultra-Human, Ultra-Smart
Meet MiniMax Speech 2.6, now live on GMI Cloud. Built for real-time voice interaction, it brings expressive, multilingual speech generation to your applications.
- Ultra-Fast: <250 ms latency for seamless real-time conversations
- Smart Text Normalization: Handles URLs, emails, dates, numbers, and more
- Full Voice Clone + Fluent LoRA: Natural, expressive, effortless delivery
- 40+ Languages: Inline code switching for smooth multilingual dialogue
Fluent anytime. In every language.
→ Try MiniMax Speech 2.6 on GMI Cloud
For Developers — Under the Hood
If you’re curious about the infrastructure powering this release, GMI Cloud’s inference stack combines containerized clusters, elastic scaling, and GPU liquidity for on-demand throughput.
Both M2 and Hailuo 2.3 run natively on GMI’s inference engine. M2 includes full support for SGLang, providing optimized token throughput and low-latency execution, while Hailuo 2.3 leverages GMI’s infrastructure for stable, high-bandwidth video generation.
Behind it all is GMI Cloud’s cutting-edge inference platform for creatives—a system architected to serve both the real-time responsiveness developers need and the high-throughput rendering pipeline creators demand.
By unifying text, code, and video inference under one scalable architecture, GMI enables seamless transitions between agentic intelligence and cinematic generation.
GMI Cloud provides deployers with configurable bare metal and container instances across regions, making it the ideal environment to experiment, fine-tune, and scale MiniMax models in production.
For open-source builders, M2 weights are available on Hugging Face; for creators, Hailuo 2.3 can be accessed through GMI’s video inference endpoint on the Console UI.
Join the Launch
The MiniMax × GMI Cloud partnership marks a turning point for AI accessibility: frontier-level models, openly available, affordable to run, and fast to deploy.
To celebrate, we’re hosting an in-house launch event in Mountain View, California, featuring live demos, hands-on workshops, and creative showcases built entirely on M2 and Hailuo 2.3.
Build AI Without Limits.
Both M2 and Hailuo 2.3 represent more than just new models—they signal the convergence of agentic intelligence and creative expression under a single platform.
- For developers, they offer new freedom to build autonomous systems at speed.
- For creators, they unlock the next medium of cinematic storytelling.
Everything is free to try during the promotion period, ending November 7th.
Start building and creating today on GMI Cloud.
.svg){kind=small}
.svg){kind=logo}
