{"id":"arete-27b","slug":"arete-27b","name":"Arete-27B","orgId":"nous-research","family":"arete-27b","releaseDate":null,"availability":"ga","openWeights":true,"license":"Apache-2.0","paramsTotal":27000000000,"paramsSource":"slug","contextWindow":null,"knowledgeCutoff":null,"modelCardUrl":"https://nousresearch.com/arete","hfRepo":null,"modalities":["text"],"aliases":["Arete-27B","arete-27b","arete 27b","NousResearch/Arete-27B","nous arete-27b","Arete"],"aliasNote":null,"origin":"tracked","tracked":{"trackedSince":"2026-11-06","announced":"2026-11-06","blurb":"Nous Research's open-weight 27B supervised fine-tune of Alibaba's Qwen3.8-27B. SFT only (~52.8M tokens, 64k context, two epochs / 1,202 steps); no RL, no preference optimization, and no architecture change. Nous frames the post-train as reinforcing Western values on a Chinese-trained base, with Inkling as the open teacher and the vision encoder frozen (text-only data). Weights on Hugging Face under Apache-2.0.","api":"none","openWeights":"yes","params":{"text":"27B","source":"Hugging Face model card / Nous blog"},"facts":[{"label":"Role","value":"General assistance and agentic work after value-alignment SFT; text generation"},{"label":"Base","value":"Qwen/Qwen3.8-27B; separate from the Qwen3.8 27B catalog page"},{"label":"Post-training","value":"Supervised fine-tuning only (~52.8M tokens, 64k context, 2 epochs / 1,202 steps). No RL, no DPO/preference optimization, no architectural modification (lab)"},{"label":"Teacher","value":"Inkling (open teacher; lab)"},{"label":"Frozen weights","value":"Vision encoder frozen; training data is text-only (lab)"},{"label":"Framing (lab)","value":"Nous describes the goal as reinforcing Western values / altering CCP-framed refusals and answers on politically sensitive topics in the base model"},{"label":"Checkpoints","value":"NousResearch/Arete-27B on Hugging Face; evaluated checkpoint step 1,202"},{"label":"License","value":"Apache-2.0"}],"claims":["Lab claim: on Nous's 803-prompt propaganda evaluation, fact mean rose from 2.32 to 2.98 (out of 5) and answers containing propaganda fell from 22% to 0.9% (Arete-27B thinking off vs Qwen3.8-27B thinking off; model card / blog).","Lab claim: Hermes Bench (internal) 65.60% for Arete-27B vs 65.77% for Qwen3.8-27B — within about 0.2 points, which Nous treats as noise (blog / card).","Lab claim: Terminal-Bench-2.1 70% for Arete-27B vs 67% for the base under Nous's like-for-like run (blog); Nous does not take credit for the difference."],"sources":[{"kind":"lab","title":"Introducing Arete-27B (Nous Research)","url":"https://nousresearch.com/arete","date":"2026-11-06"},{"kind":"repo","title":"NousResearch/Arete-27B model card (Hugging Face)","url":"https://huggingface.co/NousResearch/Arete-27B"},{"kind":"lab","title":"Nous Research on X: Arete-27B","url":"https://x.com/NousResearch/status/2118822925445050435","date":"2026-11-06"}],"firstBenchmark":null},"organization":{"id":"nous-research","name":"Nous Research","slug":"nous-research","country":null,"url":"https://nousresearch.com","color":"#11110F","mark":"N"},"price":null,"answer":{"word":"Pending benchmarks.","scope":"Announced Nov 6, 2026; no independent results yet.","lead":"As of Nov 22, 2026, Arete-27B is pending benchmarks on isitsota: announced Nov 6, 2026, with no independent results on the boards we track yet.","confidence":"low"},"status":"PENDING_BENCHMARKS","verdicts":[]}