vLLM-Omni 宣布 Day-0 支持 Runway 机器人模型 Praxis-1
Runway 把视频预训练用到了机器人控制上，Praxis-1 用 Apache 2.0 开源，vLLM-Omni 当天就支持。单张 H100 80GB 就能跑推理，还兼容 OpenPI 服务，做机器人开发的团队可以直接上手试，不用等社区适配。
Runway 发布机器人控制模型 Praxis-1，采用大规模视频预训练方式，让模型从观察世界走向在世界中执行动作。权重以 Apache 2.0 协议开源，并支持跨不同机器人本体（embodiments）进行适配，意味着同一模型可以迁移到多种硬件形态的机器人上。
推理框架 vLLM-Omni 在发布当天（Day-0）就完成了对 Praxis-1 的支持。集成方案将摄像头观测、机器人状态和任务指令作为输入，输出动作块（action chunks），同时支持离线推理和 OpenPI / Positronic 两种服务方式，并处理了会话历史与重置逻辑。整套推理经过验证，可在单张 NVIDIA H100 80GB 上运行。
这次合作由 Runway 与 Inferact 团队共同推进。对机器人开发者来说，开源权重加 Day-0 推理框架支持，大幅降低了把视频预训练模型部署到实际机器人上的门槛，单卡 H100 即可跑通也意味着中小团队可以负担得起。
🎉 Day-0 support for @runwayml’s Praxis-1 in vLLM-Omni!
From watching the world to acting in it: Praxis-1 brings large-scale video pretraining to robot control, with open weights under Apache 2.0 and adaptation across robot embodiments.
vLLM-Omni takes camera observations, robot state and a task instruction to generate action chunks. The integration supports offline inference and OpenPI / Positronic serving, with session history and reset handling. Qualified on a single @NVIDIA H100 80GB.
Thanks to the Runway and @Inferact teams for bringing this together.
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Praxis-1, Runway, vLLM, 开源模型.
Source attribution: curated from vLLM (@vllm_project). Cross-source corroboration: 11 independent sources have covered this development (including Runway ML, Jim Fan, Decoder), confirming industry-wide signal strength.
