# Arete-27B (slim)

> Arete-27B is Nous Research's fine-tune of Qwen3.8-27B, trained on 28,886 published examples so that its answers on politically sensitive topics stop following the Chinese state's account, with the base's tools, images and 262K context, under Apache 2.0.

- Full: https://www.anchorterminal.com/tools/arete-27b.md (~8,850 tokens) · this version ~1,680 tokens · JSON https://www.anchorterminal.com/tools/arete-27b.json · canonical https://www.anchorterminal.com/tools/arete-27b
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-11-10 · pricing https://www.anchorterminal.com/tools/arete-27b#pricing · reviews https://www.anchorterminal.com/tools/arete-27b#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**C · 55.1/100 · rank #105 of 135 open-weight models · #82 in Open-weight models · not agent-ready · confidence low**


Assessment: Qwen3.8-27B with its answers on politically sensitive topics retrained, four days old when read. The training data is published, which is rare. Nous Research ships BF16 only, so a small machine runs it from bartowski's GGUF, and the card documents none of the tool or thinking controls it inherits.

## Facts

- Kind: Open-weight model · vendor: Nous Research · category: Open-weight models · legal entity: Nous Research, Inc. (named in the footer of nousresearch.com) · provenance 61/100
- Packages: huggingface `NousResearch/Arete-27B`
- Auth: None · pricing: Free · x402: no · licence: Apache-2.0
- Benchmark probes: not run yet, so Performance is pending
- Surface graded: The open weights at huggingface.co/NousResearch/Arete-27B, read from the model card, the files and their metadata, the chat template, Nous Research's blog post and the runtimes' own code at their latest releases on 10 November 2026. Nothing was downloaded or run
- Model: 27,356,728,560 parameters in BF16 per the safetensors metadata, Qwen3.8-27B's architecture (Qwen3_5ForConditionalGeneration) with 64 layers and its vision encoder. The card says all language-model weights were trained, the vision encoder was frozen and the multi-token-prediction head was off during training. Text, images and video in, as the base
- Files: 18 safetensors shards in BF16 (54.7 GB), a chat_template.jinja, the tokenizer and processor configs, and an Apache 2.0 LICENSE. No pickle or Python files. Hugging Face lists a SHA-256 for every large file
- Licence: Apache 2.0, in the card's metadata and as the LICENSE file. The base, Qwen3.8-27B, is Apache 2.0 too. Nothing is added on the card
- Runtimes: The architecture is the base's, which released vLLM (0.33.0), SGLang (0.5.23) and Transformers (5.22.2) run from the safetensors files. llama.cpp b12242 converts and runs it, and Nous Research publishes no GGUF, so that route goes through bartowski's conversion. Ollama's library has no entry for it
- Context: 262,144 tokens native (max_position_embeddings in the text config). The card says training ran at 64K context
- Tools and reasoning: The chat template is the base's with small changes. It declares tools in a tools block, renders several function calls in one turn inside tool_call, and switches thinking with enable_thinking and reasoning_effort. The card documents none of this. Its evaluation ran with thinking off
- Hardware: 54.7 GB in BF16, about 66 GB with the fifth added, which fits a 128 GB unified machine. bartowski's Q4_K_M GGUF is 17.2 GB and its Q8_0 28.7 GB, each with a 0.93 GB projector
- Training data: Published. 28,886 training examples (the blog's count) from twenty thousand prompts across seven value categories, with teacher answers from Inkling beside the base model's own, about 52.8 million tokens over two epochs. The held-out evaluation, 803 prompts with a source article for each, is published too
- What changed: Answers on freedom of expression, state ideology, state versus free market, education, data privacy and surveillance, democracy and environmental policy. The card says it doesn't claim to have remedied every such behaviour

- Scores: Capability pending, Reach 50, Risk 55, Agent interface 60, Documentation and upkeep 63 · total over the 7 assessed categories
- Why:  · Reach, Memory 10, the only build from Nous Research is the 54.7 GB BF16 release, which needs about 66 GB and fits the reference machine. · Risk, File safety 25, safetensors only and no remote code. · Agent interface, Chat template 20, a Jinja file in the repository. · Documentation and upkeep, Model card 8, no limitations, context length, languages or cut-off.
- Sources: 18, open questions: 7, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/arete-27b.json
- Its vendor verifies the listing with `verify_listing` or `POST https://www.anchorterminal.com/api/v1/verify` (the full twin says how); no effect on the grade.
- Support: Discord https://discord.gg/nousresearch; https://nousresearch.com/contact; https://huggingface.co/NousResearch/Arete-27B/discussions; more in the full twin

## Before you call it

1. Start it with `llama-server -hf bartowski/NousResearch_Arete-27B-GGUF:Q4_K_M --jinja`, a third party's conversion, or serve the BF16 files with vLLM
2. Use the base's parsers in vLLM, `--reasoning-parser qwen3 --tool-call-parser qwen3_coder`, and check tool calls first, since the template differs slightly from Qwen3.8-27B's
3. Thinking switches with `enable_thinking` in the template. The card's own evaluation ran with it off
4. Expect answers on politically sensitive topics to differ from Qwen3.8-27B's. That is what the fine-tune is for
5. Validate JSON output yourself, since structured output isn't documented

## Connect

```bash
vllm serve NousResearch/Arete-27B --max-model-len 262144 --reasoning-parser qwen3 \
  --enable-auto-tool-choice --tool-call-parser qwen3_coder
# or, with llama.cpp and bartowski's GGUF, a third party's conversion:
llama-server -hf bartowski/NousResearch_Arete-27B-GGUF:Q4_K_M --jinja
```

```bash
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "NousResearch/Arete-27B", "messages": [{"role": "user", "content": "Hello!"}]}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Qwen3-Embedding-8B | BB | 75.4 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-8b.min.md |
| Gemma 4 31B IT | BB | 74.1 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-31b-it.min.md |
| Qwen3-Embedding-4B | BB | 73.6 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-4b.min.md |
| Gemma 4 12B IT | BB | 72.3 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-12b-it.min.md |
| Gemma 4 26B A4B IT | BB | 72.2 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-26b-a4b-it.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)