Persimmon v0.1 Pro/Ultra double Multi-User Turing Test score to 48% with test-time scaling
Diyi_Yang · x · 2026-11-21
Humansand launched Persimmon v0.1 Pro and Ultra tiers that use test-time scaling to make multi-turn dialogue more realistic.
- Ultra more than doubles Persimmon's score on the 8-turn Multi-User Turing Test, from 18% to 48%; at 50% a model is indistinguishable from real humans per their held-out judge
- The shared chart shows Ultra retains a 33.9% judge-fooled rate at 32 turns (Pro: 23.9%), while baselines like Nemotron Ultra Base fall to 2-6%
- Competing models start high at 2 turns (33-52%) but collapse as turns increase, highlighting Persimmon's long-conversation consistency
More from Models
- Persimmon v0.1 Pro and Ultra Now Live in Playground With API Access — gharik · 2026-11-21
- Anthropic researcher claims Claude solved General Riemann Hypothesis with 250k-page proof, unverified — inductionheads · 2026-11-21
- Hume builds safety framework from scratch for human-like user simulator models — LukeW · 2026-11-21
- Matthew Berman praises Gemini Neon: 'a great model,' explains why in video — MatthewBerman · 2026-11-21
- Microsoft launches MAI Licensed Weights; Purdue researcher asks to run them on own hardware — marlene_zw · 2026-11-21
- Out of Fable quota, dev falls back to Gemini: token rich vs token poor — jasondeanlee · 2026-11-21