Hey HN, Guanming here, cofounder of General Instinct. We just released InstinctFlash, a high-performance serving framework for robotics models, licensed under AGPL-3.0.

On Jetson Thor, we see speedups about 1.2x to 7.9x from runtime optimizations alone and up to 33.78x for LingBot-VA when combined with a distilled few-step diffusion scheduler, going from 25 visual / 50 action steps to 2 / 4 steps. Across 50 Robotwin2.0 tasks, LingBot-VA with InstinctFlash at 2/4 steps achieved a 90.5% success rate, compared with 92.1% for the baseline at 25/50 steps.

InstinctFlash supports 8 VLA / world-action model families, including pi0.5 and NVIDIA Cosmos Policy, across RTX 4090 / 5090 and Jetson Thor. Just give it your fine-tuned checkpoint and InstinctFlash handles the rest, exposing the accelerated model through a Python runtime or an OpenPI-compatible WebSocket server.

Teams at Samsung, Siemens, and other robotics startups have used InstinctFlash for model acceleration on VLAs, WAMs, and diffusion-based world models. Try it here: https://github.com/General-Instinct/InstinctFlash. Would love your feedback!