底部赞助展示区
Contact us 适合多赞助商长期展示
Loading…
适合多赞助商长期展示
MiniMax has partnered with inference platform Inferact to power production workloads for models like M3.1 Flash, with shared engineering optimizations contributed back to the open-source vLLM project.
AI developer MiniMax has established a strategic partnership with inference platform Inferact to enhance production serving for its foundation models, starting with immediate day-zero support for MiniMax M3.1 Flash. The alliance focuses on delivering reliable high-throughput serving for demanding agentic workloads under guaranteed service-level agreements.
Under this collaboration, user requests handled through MiniMax's platform can route directly through Inferact’s infrastructure alongside internal capacity. Inferact's serving stack automatically tunes endpoints against live traffic while leaving model owners in full architectural control, ensuring low latency and robust execution under peak operational conditions.
A key pillar of the agreement is the commitment to the open-source ecosystem. Technical improvements and performance patches generated from deploying MiniMax models are being contributed directly upstream to the vLLM project, enabling self-hosters and the wider community to benefit from production-grade efficiency gains.
Source review does not guarantee permanent accuracy. Consider the original source, observation time, and later corrections.
We want to thank the MiniMax team for open-sourcing MiniMax-M3, as well as MiniMax leadership for their trust and support in vLLM! The model support is led by Inferact Inc., a company aiming to grow vLLM as the world's AI inference engine and accelerate AI progress by making inference cheaper and faster. NVIDIA and AMD contributed to the hardware support. ## Related vLLM Reading MiniMax M3 builds on several areas of vLLM: [...] ### EAGLE3 Speculative Decoding MiniMax M3 also has day-0 EAGLE3 speculative decoding support in vLLM. The draft model is released at `Inferact/MiniMax-M3-EAGLE3`, enabling deployments to use a draft-model path for lower generation latency when the workload and acceptance behavior fit the target traffic. To enable EAGLE3, add a speculative decoding configuration to the serving command: ``` '{"method":"eagle3","model":"Inferact/MiniMax-M3-EAGLE3","num_speculative_tokens":3,"attention_backend":"FLASH_ATTN"}' ' ' ``` The example uses `num_speculative_tokens=3
## What each company contributes AMD is contributing access to the AMD Instinct platform, along with deep hardware and kernel expertise, AMD ROCm™ software engineering, and the AITER kernel library. Inferact and the vLLM community are contributing the engine: model integration, serving features, and day-0 enablement for new releases. The work is carried by the same engineers who run vLLM day to day and who operate the production inference traffic on Inferact's own platform, ensuring that the improvements are grounded in real production traffic. The two sides work from shared model-by-model roadmaps covering kernels, serving features, and model coverage. Everything the partnership produces will land in open source. ## Already live [...] # Inferact Collaborates with AMD to Expand Agentic Inference to vLLM on AMD Instinct™ GPUs Inferact Team Oct 2026 Announcement X min read San Francisco, October 15th, 2026 —Inferact, the company founded by the creators of vLLM to build the product
Inferact is an API launch partner for MiniMax M3.1 Flash. We're working with MiniMax on production inference for a model built to handle
Our contribution is the serving stack: bringing our vLLM engineering and the Inferact Platform to the collaboration. Thanks to the MiniMax team
ship. The engineering from both efforts is upstreamed to vLLM. Thanks to the MiniMax team for the partnership. Read the full announcement: [...] Today, we are announcing a strategic partnership with @MiniMax\_AI to advance frontier production inference for open-source models. The Inferact Platform has served MiniMax M3.1 Flash in production since the day it launched. Requests through the MiniMax platform are routed to our stack alongside MiniMax’s own capacity, after we met MiniMax’s vendor verification requirements for accuracy and reliability. The Inferact Platform serves this agentic workload with guaranteed SLAs and auto-tunes the endpoint against live traffic, while MiniMax keeps control of how its model is served. Inferact also leads day-0 @vllm\_project support for each MiniMax release, so the model serves at production quality the day the weights ship. The engineering from both efforts is upstreamed to vLLM. Thanks to the [...] Today, we are announcing a strategic partnership w
Views 1 @RyanJ_KOL @RyanJ\_KOL 46m Contributing the serving improvements back to vLLM is especially useful for self-hosters. A breakdown of which M3.1 Flash optimizations improve prefill versus decode would help teams understand where the gains come from. [...] @MiniMax_AI MiniMax (official)@MiniMax\_AI We’re partnering with Inferact to advance production inference for MiniMax models, with engineering improvements contributed back to @vllm\_project. Thanks to the Inferact team for supporting M3.1 Flash in production from day one. Looking forward to building together! @inferact Inferact@inferact 1h Today, we are announcing a strategic partnership with @MiniMax\_AI to advance frontier production inference for open-source models. The Inferact Platform has served MiniMax M3.1 Flash in production since the day it launched. Requests through the MiniMax platform are routed to our 01:31:06 AM · 10/23/2026·3,663 Views 1 @RyanJ_KOL @RyanJ\_KOL 46m [