Radar / AI infrastructure / ModelSphere 模域 (4Paradigm)
ModelSphere 模域 (4Paradigm): open-source production LLM deployment platform
4Paradigm (6682.HK) and Ruisi Zhilian launched ModelSphere on 2026-10-16 — an Apache-2.0 one-stop production AI deployment platform unifying compute management, model deployment and runtime optimization, with AutoTune scheduling and mixed domestic-GPU support (Ascend, Enflame, Moore Threads, Hygon DCU).
Why it matters
Aims at China's domestic-GPU serving gap: one control plane for heterogeneous Chinese accelerators with KV-cache-aware request routing and cluster-level autoscaling, claiming ~285% throughput gains via AutoTune plus adaptive load balancing, compatible with vLLM/SGLang engines.
What you could build with it
An enterprise team running DeepSeek or Qwen on a mixed fleet of Ascend and other domestic GPUs could use ModelSphere's unified scheduling and KV-cache-aware routing to consolidate inference clusters instead of maintaining separate per-chip deployment pipelines.
Does it hold up?
Too early to judge — announced with vendor-published throughput claims; launch articles state it is open-sourced on GitHub but give no repo link, and no docs or third-party deployments could be verified yet.
First spotted on article: source.
More AI infrastructure
DeepGEMM: DeepSeek's open GPU BLAS kernel libraryDeepSeek's DeepGEMM, a clean and efficient open-source BLAS kernel library for NVIDIA GPUs (up to 1550 TFLOPS on…infra · JEV 0.77OpenViking — self-evolving context database for AI agentsByteDance Volcengine's open-source (AGPL-3.0) context database that unifies agent memory, knowledge RAG, and skills…infra · JEV 0.76Modular open-sources MAX inference server, Mojo stdlib and accelerator kernelsThe core of Modular's unified AI deployment stack is now open: the MAX inference server (OpenAI-compatible endpoints)…infra · JEV 0.76OmniRouteFree MIT AI gateway: one endpoint, 290+ providers and 500+ models with quota-aware auto-fallback.infra · JEV 0.75Context Mode: context-window optimization for AI coding agentsMCP server for AI coding agents that sandboxes verbose tool output and persists session memory, cutting context use by…infra · JEV 0.74StrataOpen-source inference engine that runs the 125B-parameter Qwen3.8-Flash-Next MoE on a single consumer GPU with 12GB+…infra · JEV 0.74
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.