Fp8 - GoKawiil

General Instinct open-sources InstinctFlash, a fast serving engine for robotics AI models

General Instinct released InstinctFlash, an open-source inference framework that speeds up robotics 'world-action' models like LingBot-VA and pi05 on Nvidia's Jetson Thor edge computer, RTX 4090 and RTX 5090 GPUs. The company reports benchmarks showing up to a 33.78x speedup over standard PyTorch inference, achieved through FP8 quantization, reduced sampling steps, and custom acceleration kernels, with no measured drop in task performance on real robots.

github.com · 2026-09-22

DealignAI releases 'abliterated' uncensored fork of DeepSeek-V4.1-Flash

Researchers behind the dealignai project published a modified version of DeepSeek's V4.1-Flash model with its safety refusal circuitry surgically removed at the weight level, while preserving core capabilities like reasoning, vision, and its 1M-token context. The team says the checkpoint loads like a standard model with no special code needed, and claims HarmBench testing shows a 100% attack success rate for harmful prompts across both low and high reasoning effort settings, compared to the base model's much lower compliance rate.

huggingface.co · 2026-09-11