Samaya AI Releases Sterling: A 30B MoE Finance Research Agent That Outperforms Claude Fable 5.1 on FrontierFinance at 11x Lower Cost
Samaya AI has released Samaya Sterling, a post-trained model purpose-built for agentic financial research. Sterling is built on NVIDIA's Nemotron 3.5 Lightning, a 30B mixture-of-experts (MoE) model with 3B active parameters. On Samaya's FrontierFinance benchmark, Sterling scores 58.9%, edging past Claude Fable 5.1 at 55.9%. Each Sterling question costs $0.61, compared with $6.89 for Fable 5.1 โ an 11x cost advantage.
The broader takeaway for developers is a concrete recipe: co-train a small model together with its harness, and it can win decisively within a narrow domain.
TL;DR
- Size: 30B total parameters, 3B active (MoE). Base model context: up to 1M tokens. Sterling's serving context: not disclosed.
- Runs on: the base model fits a single H100 80GB in BF16 at 256K context. Sterling's serving hardware: not disclosed.
- Performance: beats every high-quality system on the FrontierFinance leaderboard while operating at a fraction of the inference cost.
via MarkTechPost