FluxVLA artık OpenAI Yanıt API'si ARACILIĞIYLA Libero'da GPT -6 Astra'yı destekliyor

FluxVLA Engine is a full-stack, end-to-end engineering platform for deploying embodied intelligence applications. Built on the core design principles of unified configuration, standardized interfaces, module decoupling, and deployability, it creates a complete engineering loop from data to real-device deployment. With the goal of providing a standardized industry–academia–research foundation, it significantly lowers the engineering barrier for VLA research and development.

[2026/09/09] 🔥 FluxVLA now supports checkpoint-free GPT-6 Astra evaluation on LIBERO through the OpenAI Responses API, including multi-view observations, native LIBERO action control, environment-configured credentials, and organized rollout artifacts.

[2026/09/09] 🔥 FluxVLA-native GR00T N1.7 is now open sourced, including training and inference workflows for LIBERO, RoboCasa, and Franka robots.

[2026/08/25] 🔥 FluxVLA-native DiT4DiT support is now available, including training and inference workflows for LIBERO and RoboCasa.

[2026/08/20] 🔥 FluxVLA-native Cosmos3 support is now available for Nano, Super, and Edge, including post-training and inference workflows.

[2026/08/13] 🔥 FastWAM world-action model support is now available.

[2026/08/10] 🔥 FluxVLA deployment on NVIDIA Jetson Orin is now supported, with edge inference acceleration reaching 7.4 Hz for GR00T-N1.5.

See orin_flashing.md for initial flashing and orin_docker_runtime.md to start the FluxVLA Docker setup.

[2026/06/30] 🔥 Single-arm and dual-arm Franka real-robot inference is now supported, including joint/eepose control configs and a deployment guide. See docs/franka.md .

[2026/06/25] 🔥 GR00T-RTC-accelerated inference is now supported, achieving 45 Hz on an RTX 5090.

[2026/06/22] 🔥 A minimal Oli humanoid whole-body (loco-manipulation) real-robot inference path (operator + runner + example config) is now available. See docs/oli_whole_body.md .

[2026/06/17] 🔥 ARM reward modeling with RA-BC/AW-BC reweighting is now supported.