Humans& Launches 744B-Param Collaborator Model Ampersand on Vercel AI Gateway

ericzelikman · x · 2026-11-17

Humans& released Ampersand, a collaborator model now live on Vercel AI Gateway (free for a week). The team claims it matches or beats frontier models like Fable and Astra in multi-user long-horizon settings on their evals, and most on external cooperation and multi-agent coding evals, at 744B params (40B active). The attached leaderboard screenshot shows Anthropic leading closed-weights at 43.5%, OpenAI at 37.8%, Google at 6.9%, with Humans& at 2.1% in 6th.

Original post →

More from coding & agent

coding & agent channel →