Persimmon v0.1 Model Card | humans&

The model card for Persimmon v0.1, a research-preview user model for human-behavior simulation, evaluation, and conversation research.

Research preview · Version 0.1 Model specs, intended uses, evaluations, limitations, and access. By humans& ai · September 10, 2026 Model Summary Field Description Model name Persimmon Model type User model Best For User understanding, simulation, evaluation, and conversation research. Base Model NVIDIA-Nemotron-3-Ultra-550B-A55B Architecture LatentMoE - Mamba-2 + MoE + Attention hybrid with Multi-Token Prediction (MTP) Total Parameters 550B (55B active) Context Length Up to 1M tokens Developer humans& ai Version v0.1 Release date 2026/09/10 Input modality Text Output…

saved by

• Timothy Kostolansky

related reading

• Today, we are releasing a research preview of our user model, along with a set of evaluations designed to measure how faithfully user models capture human behavior.persimmon.humansand.ai
• The Persona Selection Model: Why AI Assistants might Behave like Humansalignment.anthropic.com
• Surfacing Pathological Behaviors in Language Models | Transluce AItransluce.org
• Interaction Models: A Scalable Approach to Human-AI Collaboration - Thinking Machines Labthinkingmachines.ai
• [2304.03442] Generative Agents: Interactive Simulacra of Human Behaviorarxiv.org
• The persona selection model — LessWronglesswrong.com
• [2304.03442] Generative Agents: Interactive Simulacra of Human Behaviorarxiv.org
• The persona selection model \ Anthropicanthropic.com
• A Steerable Model with Emergent Capabilitiespi.website
• [2608.19549] Generating Diverse Personas for User Simulators to Test Interview Dialogue Systemsarxiv.org
• [2605.21739] AttuneBench: A Conversation-Based Benchmark for LLM Emotional Intelligencearxiv.org
• Macaron-V1-Preview: 749B MoL Agent Model post-trained from GLM5.1macaron.im