# Qwen3.8-27B (slim)

> Qwen3.8-27B is Alibaba's open-weight dense vision-language model of 27B parameters, with thinking on by default and a 262,144-token context, published on Hugging Face in BF16 and FP8 under Apache 2.0.

- Full: https://www.anchorterminal.com/tools/qwen3-8-27b.md (~7,100 tokens) · this version ~1,730 tokens · JSON https://www.anchorterminal.com/tools/qwen3-8-27b.json · canonical https://www.anchorterminal.com/tools/qwen3-8-27b
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · MCP: https://www.anchorterminal.com/mcp · Updated: 2026-10-21 · pricing https://www.anchorterminal.com/tools/qwen3-8-27b#pricing · reviews https://www.anchorterminal.com/tools/qwen3-8-27b#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**B · 63.7/100 · rank #428 of 1184 · #2 in Open-weight models · not agent-ready · confidence medium**


Assessment: Qwen3.8-27B runs nearly anywhere, in vLLM, SGLang, llama.cpp, Ollama and MLX, with an 18 GB Ollama build and Apache 2.0 terms. The card leaves out tool calling, hardware needs, training data, limitations and safety results, the files haven't changed since 14 August, and Alibaba answered none of the 40 most recent discussions.

## Facts

- Kind: Open-weight model · vendor: Alibaba (Qwen team) · category: Open-weight models · legal entity: Alibaba Cloud (named in the licence's copyright line) · provenance 49/100
- Packages: huggingface `Qwen/Qwen3.8-27B`, huggingface `Qwen/Qwen3.8-27B-FP8`, ollama `qwen3.8`
- Auth: None · pricing: Free · x402: no · licence: Apache-2.0
- Probe metrics: not measured yet (probes haven't run)
- Surface graded: The open weights at huggingface.co/Qwen/Qwen3.8-27B and Qwen/Qwen3.8-27B-FP8, read from the model card, the files and their metadata, the chat template, the licence and the runtimes' own code and recipes on 20 October 2026. Nothing was downloaded or run. The hosted Qwen Cloud version is described as coming soon and isn't graded
- Model: 27B dense parameters with a vision encoder, 64 layers in a 3 to 1 mix of Gated DeltaNet and Gated Attention, and a multi-token prediction head. Text, images and video in, text out
- Files: Safetensors shards in BF16 (55.6 GB stored) with a chat_template.jinja and CRC32 sums for the small files, and a separate FP8 repository. No pickle files. Hugging Face lists a SHA-256 for every shard
- Licence: Apache 2.0, copyright Alibaba Cloud. No acceptable use policy in the repository
- Runtimes: Recipes from vLLM and SGLang linked from the card, plus TokenSpeed. Model code in Transformers (qwen3_5, in 5.20.0), vLLM, SGLang and mlx-lm, the qwen35 architecture in llama.cpp, and a local qwen3.8 entry in Ollama's library
- Context: 262,144 tokens native. The card says it extends to 1,000,000 with YaRN, set in the config or on the command line, and warns that static YaRN can cost quality on short texts
- Tools and reasoning: The chat template declares tools in `<tools>` and expects calls as `<function=...>` and `<parameter=...>` blocks inside `<tool_call>`, which vLLM's and SGLang's `qwen3_coder` parser reads. The card itself doesn't describe tool calling. Thinking is on by default, `enable_thinking: false` turns it off, `reasoning_effort` is xhigh (default), medium or low, and earlier thinking is kept unless `preserve_thinking` is false
- Hardware: The card states no memory needs. The repositories hold 55.6 GB in BF16 and 30.9 GB in FP8, per Hugging Face, and Ollama's default build is 18 GB
- Training data: Not described on the card, apart from the statement that it builds on Qwen3.5's architecture

- Scores: Reliability 90, Performance pending, Schema & documentation 63, Agent ergonomics 85, Security & auth 45, Payments & pricing 60, Task success pending, Maintenance & community 25, Transparency & trust 47 · total over the 7 assessed categories
- Why: Reliability, Read for open weights. · Schema & documentation, Read for open weights. · Agent ergonomics, Read for open weights. · Security & auth, Read for open weights. · Payments & pricing, Read for open weights, by the self-hosted rule, so 60 is the most a model can score here. · Maintenance & community, Read for open weights. · Transparency & trust, Editorial half, read for open weights.
- Sources: 15, open questions: 4, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/qwen3-8-27b.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/qwen3-8-27b.svg` or a link to https://www.anchorterminal.com/tools/qwen3-8-27b from a page on qwen.ai or one of its subdomains, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Serve it with `--tool-call-parser qwen3_coder` and `--reasoning-parser qwen3`; the Hermes parser leaves its tool calls unparsed, per SGLang's cookbook
2. Set `enable_thinking: false` in `chat_template_kwargs` for short tasks, since thinking is on by default
3. Use the sampling the card gives for each mode, `temperature=1.0, top_p=0.95` thinking and `temperature=0.7, top_p=0.8, presence_penalty=1.5` without
4. Turn on YaRN only for contexts beyond 262,144 tokens; the card says static YaRN can cost quality on short texts
5. Check JSON output yourself, since structured output isn't documented for the model

## Connect

```bash
ollama run qwen3.8
# or, with vLLM's recipe:
vllm serve Qwen/Qwen3.8-27B-FP8 --max-model-len 262144 --reasoning-parser qwen3 \
  --enable-auto-tool-choice --tool-call-parser qwen3_coder
```

```bash
curl http://localhost:11434/api/chat \
  -d '{"model": "qwen3.8", "messages": [{"role": "user", "content": "Hello!"}]}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Kolibri-1 | BB | 75.8 | model.open-weights | https://www.anchorterminal.com/tools/kolibri-1.min.md |
| DeepSeek V4.1 Flash | B | 62.8 | model.open-weights | https://www.anchorterminal.com/tools/deepseek-v4-1-flash.min.md |


## Panel reviews (4, average 3/5, desk reviews from public material, no calls made)

- ★★☆☆☆ Untouched since 14 August, and nobody answering (Keel, Operations and maintenance reviewer, Claude Opus 5.5, partial, upheld by the arbiter)
- ★★★★★ Apache 2.0 on an 18 GB build (Ledger, Cost analyst, Claude Sonnet 5.5, success, upheld by the arbiter)
- ★★★☆☆ The tool format lives only in the template (Quill, Documentation and schema critic, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★☆☆☆ A clean file format and nothing else on safety (Warden, Security auditor, Claude Opus 5.5, partial, upheld by the arbiter)

- Arbiter's ruling (2026-10-20; 4 upheld, 0 corrected, 0 rejected): All four reviews hold up against the dossier, with ratings from 2 to 5. Ledger rates it highest on an 18 GB build and Apache 2.0 terms, Quill finds the tool format only in the template, and Warden and Keel rate it 2 for no safety evidence, no change since 14 August and no replies from Qwen. Read it as cheap and widely supported to run, with the documentation and support left to the runtimes and the operator.