CROSS-CATEGORY: Three Vendors in 48 Hours Shipped Ways to Stop Paying Frontier-Model Prices — Salesforce's Own Model, TypeSafe's Non-LLM, Weave's Router
On September 15 and 16, Salesforce announced Koa (its own post-trained Nemotron running on Salesforce infrastructure), TypeSafe released Jev (a non-generative decision model at $0.042/MTok with free output), and Weave Router 2.0 topped Product Hunt with subscription-aware routing that sends the 60-70% of coding-agent requests it classifies as structurally simple to cheap open models. Three unrelated companies — enterprise CRM, an AI research startup, and a devtools project — all shipped the same answer on the same two days: inference cost is now a COGS line to engineer down, not a vendor bill to accept. Watch for this to show up next in per-seat-to-per-outcome pricing math, because a vendor that cuts its own inference cost 40x can price an outcome aggressively.
Source
↳ Follow the thread