Ling-3.0-flash-fin
InclusionAI (Ant Group) · China · 2026
A finance-tuned sibling of Ling-3.0-flash that appeared at hosting providers on 27 August 2026 — and, unlike the rest of the family, without weights or a model card.
Ling-3.0-flash-fin is a finance-specialised variant of Ling-3.0-flash, the current open-weight generation from InclusionAI, the open-source arm of Ant Group. It carries the same mixture-of-experts body as the base model — 124 billion total parameters with about 5.1 billion activated per token — and the same 262,144-token context window, with output capped at 32,768 tokens per response. What makes this entry unusual is what is missing. Every other Ling-3.0 release of 2026 arrived as downloadable weights on Hugging Face and ModelScope, most of them under an MIT declaration. The finance variant did not. As of 1 September 2026 there is no repository for it in the inclusionAI organisation, no model card, no licence file and no announcement in the publisher's own repositories or cookbook. The model exists only as an endpoint at commercial hosting providers, which is where every published figure about it comes from. Those providers describe it as built for real investment work: multi-step financial tasks, document analysis and longer-horizon planning, with the base model's general reasoning, coding and mathematics retained. No benchmark results, training data description or evaluation on financial tasks has been published by anyone, including the maker. Readers should treat the finance claim as a vendor description rather than a measured result. Access is currently free of charge. OpenRouter lists it only as a zero-priced endpoint and Novita prices both input and output at zero, while the ordinary Ling-3.0-flash on the same platform costs money. A free tier with no weights and no paper is a familiar pattern for a model being trialled in public before a commercial decision, and the terms can change without notice. This catalogue keeps the entry separate from Ling-3.0-flash because the two differ in the way that matters most to a reader: one can be downloaded and run, the other cannot.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!