Ling 3.0 Flash Fin: a free, finance-tuned agent live on OpenRouter
Built on Ant Group's open-weights Ling 3.0 Flash, the 124B MoE (5.1B active) is tuned for long-horizon investment workflows and agentic tool use.
Copy markdown
A finance agent you can call for free today
InclusionAI put Ling 3.0 Flash Fin on OpenRouter on Aug 27, free via Novita — a 124B MoE with only 5.1B active params, a 256K context window, and 32K max output, plus function calling and reasoning. It's tuned for multi-step, long-horizon investment workflows while keeping general coding and math.
The base model is genuinely strong — and MIT-licensed
Fin sits on Ling 3.0 Flash, which Artificial Analysis rates at Intelligence Index 38 — #1 among open-weights models its size (median 9) and on the Pareto frontier. The general weights are MIT on Hugging Face (BF16 ~255GB, FP8 ~128GB), and the first-party API runs about $0.075/$0.22 per 1M tokens.
Weekend build: a filings-to-memo agent
A 256K context (Ant says the architecture scales toward 1M) plus function calling and long-horizon planning make it a cheap backbone for an agent that ingests 10-Ks and earnings calls and drafts an investment memo — prototype it free before paying for a frontier model.
The catch: Fin's weights aren't public
Only the general Ling 3.0 Flash is open; the finance-tuned Fin is API-only for now, and "free" runs through the OpenRouter/Novita promo, not a permanent price. Both are text-only — no image input.