Ling 3.0 Flash Sante: a free, medical-tuned 124B model on OpenRouter
inclusionAI now serves free health- and finance-tuned versions of its 124B MoE, plus a $0.021 base model — cheap domain agents you can wire up this weekend.
Copy markdown
Sante: a free clinical reasoner, live now
Ling 3.0 Flash Sante landed free on OpenRouter Sept 4 — a 124B MoE (5.1B active, 262K context) tuned for medical knowledge reasoning, clinical safety, and evidence-based retrieval. You can prototype a triage or medical-Q&A assistant at zero token cost.
Fin: the finance sibling, also $0
The finance-tuned Ling 3.0 Flash Fin (Aug 27) targets investment workflows and long-horizon planning while keeping general coding and math. Same 124B/5.1B core, 262K context, free via NovitaAI — two domain specialists for nothing.
The base model is almost free too
Want a generalist instead? Base Ling 3.0 Flash runs $0.021 in / $0.063 out (~65% off), with hybrid reasoning that scales its thinking to task difficulty and a 262K window — built for production-scale agentic inference on a budget.
The catch: one provider, thin scores
All three route only to NovitaAI, so there's no fallback if it wobbles, and the Fin variant posts just 19.7% on GPQA Diamond. Treat these as narrow domain tools, not frontier generalists, and don't make them a single point of failure.
Elsewhere: IBM's Granite 4.2 8B
IBM's dense 8B reasoner (Aug 31) lists at $0.06 in / $0.25 out with a 131K window, 12 languages, and full/low/no-thinking modes — a cheap, local-friendly agent workhorse with weights on Hugging Face.