San Francisco, October 23, 2026 — MiniMax, a leading developer of frontier AI models, and Inferact, founded by the creators of vLLM, today announced a strategic engineering partnership to make open-weight AI models easier to deploy and scale in production.
The partnership focuses on two areas: delivering optimized day-zero vLLM support for new MiniMax model releases and advancing production inference through joint engineering and real-world deployment. Inferact Platform has served MiniMax M3.1 Flash production traffic since launch, with engineering improvements and operational learnings contributed back to the open-source vLLM ecosystem.
Day-Zero Support for Frontier Open Models
Making frontier models openly available is only the first step. Developers and enterprises also need reliable, efficient inference infrastructure to deploy them at scale.
MiniMax's models, with capabilities spanning long-context reasoning, multimodal intelligence, and large mixture-of-experts architectures, introduce increasingly complex serving requirements. MiniMax and Inferact work together ahead of releases to optimize and validate model-specific serving paths in vLLM, targeting production-ready support from the day new weights become available.
This collaboration includes compatibility testing, decoding optimizations, and performance validation across supported hardware platforms. Engineering improvements are upstreamed to vLLM, benefiting the broader developer community.
Production Inference for Agentic Workloads
Inferact Platform has served MiniMax M3 and M3.1 Flash in production since launch, supporting traffic through the MiniMax platform alongside MiniMax's own infrastructure. The deployment was validated against MiniMax's requirements for model quality, performance, and reliability.Designed for agentic workloads including multi-step automation, software engineering, and visual reasoning, M3.1 Flash requires efficient handling of long contexts, repeated tool interactions, and sustained performance across complex workflows.
Combining MiniMax's model expertise and production workload insights with Inferact's inference engineering, the teams optimize serving efficiency and reliability. Inferact Platform provides production-grade inference built around vLLM, with service-level guarantees and automated performance tuning as workloads evolve. Improvements validated through real-world traffic also inform further vLLM optimizations.
Advancing Open-Source Inference
As frontier models evolve, inference infrastructure must keep pace with new architectures, larger context windows, multimodal capabilities, and increasingly demanding agentic workloads.
Through joint engineering and production validation, MiniMax and Inferact bring model-specific optimizations into vLLM, improving compatibility, decoding efficiency, and performance across supported hardware platforms. These contributions help developers and enterprises deploy MiniMax models more efficiently on their own infrastructure.
The companies plan to deepen their collaboration on day-zero vLLM support, production inference optimization, and open-source contributions for future MiniMax model releases, building on their existing work with M3.1 Flash.
About MiniMax
MiniMax is a leading global artificial intelligence company with a mission of “Intelligence with Everyone.” The company develops its own general-purpose foundation models across text and multimodal intelligence and brings them to users worldwide through AI-native products and an Open Platform for enterprises and developers. Today, MiniMax's models and products serve more than 300 million users across over 200 countries and regions, and more than one million enterprises and developers. For media inquiries, please contact pr@minimax.io
About Inferact
Inferact is building the production inference infrastructure for AI. Founded by the creators and core maintainers of vLLM, the world's most widely adopted open-source LLM inference engine, Inferact makes it faster, cheaper, and easier to serve frontier AI models at scale, across models and hardware, on infrastructure companies control.
Media Contact: contact@inferact.ai
