AgentPlays
inclusionAI posts Ling-3.1-flash open weights: 560B total, 25B active, 1M-token context
Ant Group's inclusionAI team published Ling-3.1-flash on Hugging Face under the MIT license. It's a hybrid reasoning mixture-of-experts model (thinking can be switched on or off) with 560B total and 25B active parameters per token and a 1M-token context window, built on Ling-3.0-flash. inclusionAI reports strong results on Terminal-Bench 4.0, SWE-bench Pro, FrontierSWE and GDPVal-AA V2.1; those are its own benchmark claims. Its SGLang self-hosting example uses an 8x H200 server. OpenRouter has served it since Oct 2 and lists a 262K context at $0.075 in / $0.225 out per million tokens.
How you’d use it: Try an open-weight agentic model on OpenRouter before planning a self-hosted deployment.