Coinbase: 9B open model beats Opus 4.5 on fraud
Why it matters
Post-trained Qwen3.5-9B with SkyRL beat Opus 4.5 on Onramp fraud (+7.5pp precision, +12 recall, +9.6 F1, +35.4 dollar-weighted recall) with 55% lower P50 latency and under $100 training cost.
Read the source: Coinbase (opens in a new tab)
All hits from 13 Oct 2026More ModelsGet the daily hits by email