worth noting the trend here --> @harvey post-trained an open weight model for legal work + now @coinbase for fraud narrow task + proprietary…

worth noting the trend here --> @harvey post-trained an open weight model for legal work + now @coinbase for fraud narrow task + proprietary data + open weights seems to beat frontier on cost & speed (tho coinbase’s comp was vs Opus 4.5 tbf) fraud is a GREAT use case for this --> super low tolerance re: latency, chargebacks + losses tell you what was fraud so u get training data for free, + it's data you’d prolly never send to a lab anyway once again raises the question --> can vertical ai cos (+ big cos in-house) build models faster than the labs can build vertical products?? We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models, at much lower cost and speed, for a specialized task (fraud prevention in this case). I think we'll see more of this pattern inside compani…

We were able to fine tune an open weight model (Qwen3.5-9B) on proprietary internal data at @coinbase, and it out performed frontier models,…
Two bits here I think we're going to see a lot more: 1. custom harnesses / finetunes on smaller open source models to beat frontier models a…
Harvey just published a study showing a hybrid setup, open source GLM 5.1 as primary worker, routing to Opus 4.7 only when needed beats pure…
Routing and post-training open-source models won't only give you more accurate systems but also meaningfully faster and cheaper systems as m…

Source: Clem Delangue (X) | 2026-10-12

Related