Samaya releases Sterling, a 3B-active finance research model
Original title: Samaya AI Releases Sterling: A 3B-Active Finance Research Agent Model that Beats Claude Fable 5.1 on FrontierFinance at 11x Lower Cost
On Samaya’s own benchmark it scores 58.9% at $0.61 per question versus 55.9% and $6.89 for Claude Fable 5.1.
Samaya AI released Sterling, a LoRA post-train of NVIDIA’s Nemotron 3.5 Lightning for agentic financial research. The base is a 30B mixture-of-experts model with 3B active parameters. Samaya says Sterling scores 58.9% on FrontierFinance, above Claude Fable 5.1 at 55.9%, and costs $0.61 per question versus $6.89. No public weights or API were announced. All FrontierFinance figures are self-reported.
Unconfirmed: Scores, costs, and the claim that benchmark questions were excluded from training are Samaya’s self-reported figures.
- High-effort mode reaches 60.6% at $1.83 and 372 seconds per question.
- Training used two supervised stages and offline then online GRPO on Tinker.
- The hardest third of questions scored 45.2%, up from 31.8%.
- Samaya says no benchmark question was used in training.
- The model itself costs $0.04 per question; tools and helpers make up the rest.
- Organizations
- Thinking Machines Lab
From the original article
7 min read at marktechpost.comSamaya AI’s Sterling is a post-trained Nemotron 3. 5 Lightning model for financial research. It scores 58.
By Asif Razzaq · marktechpost.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.