Ling-3.1-flash

Developers can run the model for reasoning, coding, and tool-using agent tasks using SGLang or vLLM.

Availability

Available on Hugging Face and ModelScope; also listed on OpenRouter

Pricing model

Not established

Deployment

Inference instructions provided for SGLang and vLLM

Privacy

Not established

Licence

MIT

• The model has 560B total parameters and 25B active parameters per token. ^[1]

  With **560B** total parameters, **25B** active parameters per token, and a **1M-token context window**, Ling-3.1-flash builds upon Ling-3.0-flash with improvements in architecture and performance on real-world tasks.

• The model has a 1M-token context window. ^[1]
• The model is designed for general-purpose agents, search, office productivity, and software engineering. ^[1]

  With stronger real-world capabilities, Ling-3.1-flash shows improvements across a range of real-world tasks, including general-purpose agents, search, office productivity, and software engineering.

• The model is available on Hugging Face and ModelScope.
• The release provides inference instructions for SGLang and vLLM. ^[1]

  Verified recipes for H200, GB300, H800, H100, and B200, plus HiCache (Mooncake L3), reasoning on/off, tool calling, and measured accuracy/speed numbers, are in the [SGLang Ling-3.1-flash cookbook](https://docs.sglang.io/cookbook/autoregressive/InclusionAI/Ling-3.1-flash).

• Extract structured text from PDFs and images on a CPU Daily brief 2026-10-14