The model has immediate support from the vLLM and SGLang inference frameworks for serving on private infrastructure. By applying MXFP4 and block-FP8 quantization, SGLang reduces the weights of the model from 534GB to 292GB, enabling the system to fit on 4 Hopper GPUs.
Wednesday, October 21, 2026
Story
AntLingAGI Open-Sources 560B Parameter Ling-3.1-flash Model
AI7 days ago
Developers can now generate interactive mobile UIs and web pages from rough briefs using a new open-weights reasoning model. AntLingAGI released Ling-3.1-flash, a 560B parameter mixture-of-experts model with 25B active parameters per token and a 1M token context window, to target coding, office, and agent tasks. The model's performance has been benchmarked on GDPVal-AA, FrontierSWE, and Terminal-Bench 4.0.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.