# Ant Group's Ling-3.1-flash is open weights under MIT

> inclusionAI published Ling-3.1-flash on Hugging Face under MIT: 560B total parameters, 25B active, and a 1M-token context on the model card.

Published 2026-10-14T08:00:00-07:00 · Daily Digest · Canonical: https://news.littlewolflabs.com/posts/2026-10-14-ling-3-1-flash-open-weights

[inclusionAI](https://huggingface.co/inclusionAI/Ling-3.1-flash), Ant Group's AI lab, published Ling-3.1-flash on Hugging Face under MIT. Mixture-of-experts, 560B total / 25B active parameters, 1M-token context per the model card. Card benchmarks are inclusionAI's own: Terminal-Bench 4.0, SWE-bench Pro, FrontierSWE, GDPVal-AA. X posts highlight website, UI and 3D-prototype generation, but those are early impressions.

## What it costs

Weights free, but recommended server is 8x H200, not a home-PC model. On [OpenRouter](https://openrouter.ai/inclusionai/ling-3.1-flash), DeepInfra lists $0.30/M input, $0.90 output; Novita shows $0.075/$0.225 with a 75% discount applied, both at 262K context.

## Should you switch

Cheap API option worth testing for coding agents and UI prototypes; run locally only with datacenter GPUs.