# Arete-27B

Arete-27B is Nous Research's latest language model in our dataset, released Nov 6, 2026. Open weights (Apache 2.0). Nous Research's Apache 2.0 supervised fine-tune of Qwen3.8-27B (27B) aimed at reducing state-aligned framing; released on Hugging Face Nov 6, 2026. Verified against vendor sources on Nov 8, 2026.

Primary source (source of record; Agentic Weekly is the fast independent index): https://huggingface.co/NousResearch/Arete-27B

Last verified: 2026-11-08 · Evidence: Reported · Canonical: https://theagenticweekly.com/models/arete-27b · JSON: https://theagenticweekly.com/api/models/arete-27b

## Key facts

| Fact | Value | Source | Label | Anchor |
|---|---|---|---|---|
| Vendor | Nous Research |  | editorial | https://theagenticweekly.com/models/arete-27b#fact-vendor |
| Released | Nov 6, 2026 | https://huggingface.co/NousResearch/Arete-27B | vendor-reported, confirmed | https://theagenticweekly.com/models/arete-27b#fact-release-date |
| Parameters | 27B total | https://huggingface.co/NousResearch/Arete-27B | vendor-reported, confirmed | https://theagenticweekly.com/models/arete-27b#fact-parameters |
| Licence | Apache 2.0 (open weights) | https://huggingface.co/NousResearch/Arete-27B | vendor-reported, confirmed | https://theagenticweekly.com/models/arete-27b#fact-license |

## Vendor claims and neutral signals

- Nous Research reports answers with propaganda falling from 22% (Qwen3.8-27B) to 0.9% on its 803-prompt eval, with Hermes Bench 65.60% vs 65.77%. These are vendor figures.

## What we tested

Nothing yet. We have not run Arete-27B ourselves, so every fact here is Reported from the sources below.

## Not verified, left out

- Inference context window and API pricing were not stated on the model card (training context was 64k).
- All benchmark figures are Nous Research's own runs, including an LLM-judged internal propaganda evaluation.

## Alternatives to Arete-27B

Models from other vendors that share at least one of our groups, from each model's sourced facts (not tested by us).

- [Ampersand v0.1](https://theagenticweekly.com/models/ampersand-v0-1) (Humans&): Both in: Agents · released Nov 6, 2026
- [GPT-6 Cyber](https://theagenticweekly.com/models/gpt-6-cyber) (OpenAI): Both in: Agents · $20 in / $100 out per 1M tokens · 1.05M context · released Nov 5, 2026
- [Inkling 2 Small](https://theagenticweekly.com/models/inkling-2-small) (Thinking Machines Lab): Both in: Agents · released Nov 5, 2026
- [LongCat-2.5](https://theagenticweekly.com/models/longcat-2-5) (LongCat, Meituan): Both in: Agents · $0.3 in / $1.2 out per 1M tokens (launch promo, Listed as a limited-time discount with no end date or regular price given.; list price after the promo not stated) · 1M context · released Nov 2, 2026
- [Gemini 4 Carbon](https://theagenticweekly.com/models/gemini-4-carbon) (Google): Both in: Agents · $2 in / $10 out per 1M tokens (introductory price, end date not stated; list $4 / $20) · released Oct 28, 2026
- [GPT-6.1 Terra](https://theagenticweekly.com/models/gpt-6-1-terra) (OpenAI): Both in: Agents · $1 in / $5 out per 1M tokens · 1.05M context · released Oct 26, 2026
- [Best AI model for agents (Nov 2026): Claude Opus 5.5 leads Terminal-Bench 4.0](https://theagenticweekly.com/best/ai-model-for-agents)
- [Best open-weight LLMs (Nov 2026): GLM-5.3 leads Terminal-Bench 4.0](https://theagenticweekly.com/best/open-weight-llms)

## Arete-27B FAQ

**When was Arete-27B released?**

Nous Research released Arete-27B on Nov 6, 2026. Source: https://huggingface.co/NousResearch/Arete-27B

**Is Arete-27B open weights?**

Yes. The weights are published under Apache 2.0. Source: https://huggingface.co/NousResearch/Arete-27B

**Has Agentic Weekly tested Arete-27B?**

Not yet. Every fact on this page is Reported: read from the linked vendor pages, not measured by us.

## Sources

- [Arete-27B model card (Hugging Face)](https://huggingface.co/NousResearch/Arete-27B)
- [Agentic Weekly brief: Arete-27B](https://theagenticweekly.com/news/nous-arete-27b)

Data licence: CC BY 4.0. Data: Agentic Weekly (https://theagenticweekly.com), CC BY 4.0.