Edge0
Run large mixture-of-experts models on a Mac by streaming experts from SSD
Emerged finds new products worth a look, daily — what’s new today
InstinctFlash
A serving runtime that speeds robot policy models up for real control loops
Robotics teams whose VLA or world-action models are too slow on the robot can point this at a fine-tuned checkpoint and get measured speedups.
InstinctFlash from General Instinct accelerates eight robotics model families, including pi0.5, GR00T, LingBot and Cosmos policies, on Jetson Thor, RTX 4090 and RTX 5090. It combines CUDA graphs, cache reuse, tailored kernels, FP8 and optional few-step distillation, and serves through a Python runtime or an OpenPI-compatible WebSocket. Published benchmarks show 1.2x to 7.9x from runtime work and up to 33.78x on LingBot-VA, with success on 50 Robotwin tasks at 90.5% against a 92.1% baseline. Reproduction scripts are included.
More like this
Warp
Run trillion-parameter models on a laptop by streaming weights from NVMe
Goethe
Turn a Linux box with NVIDIA GPUs into a private, OpenAI-compatible endpoint
Jinfer
An AI inference engine that runs inside the JVM, with no Python sidecar
Strata
Runs a 125B mixture-of-experts model on a single gaming PC, fully offline
Reef
Open-source infrastructure for agents that keep improving from use