by InclusionaiReleased 4 September 2026
A health and medicine mixture-of-experts model from InclusionAI's Ling 3.0 line with 5.1B active parameters of 124B, reasoning before it answers and calling your tools on clinical text
Reads about 260 pages · writes up to 32 pages
Standard OpenAI fields, grouped by what they control. Anything not listed is accepted and ignored.
Shaping the reply
How long, how varied, where to stop.
max_tokenstemperaturetop_ptop_kstopseedfrequency_penaltypresence_penaltyrepetition_penaltylogprobstop_logprobsReasoning
Let it think before answering — and see the thinking.
reasoninginclude_reasoningTools
Let the model call your functions.
toolstool_choiceNo. Inclusionai has retired it. Other Inclusionai models are on sale.
A health and medicine mixture-of-experts model from InclusionAI's Ling 3.0 line with 5.1B active parameters of 124B, reasoning before it answers and calling your tools on clinical text.
Inclusionai names no languages for Ling 3.0 Flash Sante. Arabic, French and Darja: not stated.
It reads text. Up to 262K tokens in one request, about 260 pages. It writes up to 25 000 words at once.
No. Inclusionai has not published its weights.
Inclusionai. Released 4 September 2026.
Ling 3.0 Flash is a text model from inclusionAI, built to finish automated tasks using less text, which keeps cost and waiting down. You send up to about 262 pages of text, with tools if you wish, and get text back. Use it for high-volume automated tasks where the amount of text used and speed set your limits.
Ling 3.1 Flash is a text model from inclusionAI that mixes quick answers with thinking first. It reads up to about 262 pages at once, can write up to about 236 pages in a single reply, and can use tools or fill in an exact form you give it. Pick it when one request needs a very long answer.
Ling 3.0 Flash VL reads text, pictures and video and answers in text, so you can send a screenshot, a photo or a clip along with your question. It can switch between an instant answer and thinking first, and it can use tools, which suits automated helpers that must act on what is on a screen. It reads about 131 pages at once.
This model processes text input up to 260 pages and produces text output up to 250 pages. It is best suited for complex instruction following and managing multi-turn conversations for automated workflows. A key feature is its ability to act as a responsive agent and utilize provided tools. However, it does not have the capability to search the web or access external information.
From what its maker and two benchmark sites publish, and where it stands on OpenDunes. Read 13 Oct 2026.
Thinks first
Works through the problem before it answers.
Uses your tools
Can call functions your code provides.
Fills your JSON No.
Answers in the JSON shape you give it.
Searches the web No.
Can look things up online while it answers.
Cheaper on repeats
A prompt it saw a moment ago costs less the second time.
Reads long clinical records and medical texts in one pass up to 262144 tokens
Thinks first
Calls your tools inside health workflows and reasons over what they return
Thinks first · Uses your tools