{
 "slug": "Trinity_(model_family)",
 "title": "Trinity (model family)",
 "type": "family",
 "short_desc": "Arcee AI's open-weight sparse mixture-of-experts line: edge-scale Nano, agent-focused Mini, the 400B Trinity Large, and a Thinking reasoning branch.",
 "categories": [
  "Model families",
  "Open-weight models",
  "Mixture-of-experts models",
  "Arcee models"
 ],
 "infobox": {
  "Developer": "Arcee AI",
  "Debut": "Trinity Mini and Trinity Nano Preview, December 1, 2025",
  "Architecture": "Sparse mixture-of-experts Transformer",
  "Scales": "Nano (6B total / 1B active), Mini (26B / 3B), Large (400B / 13B)",
  "License / access": "Apache 2.0 (developer-confirmed for Trinity Large Thinking, per launch coverage for earlier releases); weights on Hugging Face"
 },
 "infobox_links": {
  "Developer": [
   "Arcee_AI"
  ],
  "Architecture": [
   "Mixture_of_experts",
   "Transformer_(architecture)"
  ],
  "Scales": [
   "Trinity_Nano",
   "Trinity_Mini",
   "Trinity_Large"
  ],
  "License / access": [
   "Hugging_Face"
  ]
 },
 "aliases": [
  "Trinity"
 ],
 "family": "",
 "org": "Arcee AI",
 "registry_status": "stub",
 "words": 867,
 "references": 7,
 "outbound": [
  "AFM-4.5B",
  "Arcee_AI",
  "Claude_Opus_4.6",
  "DeepSeek-R1",
  "DeepSeek_(model_family)",
  "Distributed_training",
  "GLM-5",
  "GLM_(model_family)",
  "Hugging_Face",
  "Instruction_tuning",
  "Kimi_(model_family)",
  "Kimi_K3",
  "Knowledge_distillation",
  "Large_language_model",
  "Llama_(model_family)",
  "Llama_3.1",
  "Llama_4",
  "MMLU",
  "Meta_AI",
  "Mixture_of_experts",
  "NVIDIA",
  "Scaling_laws",
  "Supervised_fine-tuning",
  "Test-time_compute",
  "Transformer_(architecture)",
  "Trinity_Large",
  "Trinity_Large_Thinking",
  "Trinity_Mini",
  "Trinity_Nano",
  "o1"
 ],
 "inbound": [
  "AFM-4.5B",
  "Arcee_AI",
  "Mixture_of_experts",
  "Trinity_Large",
  "Trinity_Large_Thinking",
  "Trinity_Mini",
  "Trinity_Nano"
 ],
 "url": "wiki/Trinity_(model_family).html",
 "built": "2026-07-24"
}