# Qwen3.6-35B-A3B (slim)

> Qwen3.6-35B-A3B is Alibaba's open-weight mixture-of-experts vision-language model, 35B parameters with 3B active and a 262,144-token context, published on Hugging Face in BF16 and FP8 under Apache 2.0.

- Full: https://www.anchorterminal.com/tools/qwen3-6-35b-a3b.md (~12,600 tokens) · this version ~2,130 tokens · JSON https://www.anchorterminal.com/tools/qwen3-6-35b-a3b.json · canonical https://www.anchorterminal.com/tools/qwen3-6-35b-a3b
- 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/qwen3-6-35b-a3b#pricing · reviews https://www.anchorterminal.com/tools/qwen3-6-35b-a3b#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**B · 63.2/100 · rank #25 of 100 open-weight models · #20 in Open-weight models · not agent-ready · confidence high**


Assessment: A fast mixture-of-experts cousin with 3B active parameters for coding and tool calling, with image and video input, in every major runtime under plain Apache 2.0. Ollama's 4-bit build needs about 28 GB. The main risk is that nobody at Alibaba maintains it, with no safety results, no security contact and no replies to tool-calling reports.

## 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/Qwen3.6-35B-A3B`, huggingface `Qwen/Qwen3.6-35B-A3B-FP8`, ollama `qwen3.6:35b`
- 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/Qwen3.6-35B-A3B and Qwen/Qwen3.6-35B-A3B-FP8, read from the model card, the files and their metadata, the chat template, the licence, the publisher's GitHub repository and the runtimes' own code and recipes on 21 October 2026. Nothing was downloaded or run. The hosted version on Alibaba Cloud Model Studio isn't graded
- Model: 35B parameters in total (35,951,822,704 in the BF16 files, vision encoder included) with 3B active, 40 layers laid out as 10 blocks of three Gated DeltaNet layers and one Gated Attention layer, each followed by a mixture of 256 experts with 8 routed and 1 shared active, and a multi-token prediction head. Text, images and video in, text out
- Files: 26 safetensors shards in BF16 (71.9 GB) with a chat_template.jinja, and a separate FP8 repository (37.5 GB). No pickle files. Hugging Face lists a SHA-256 for every shard
- Licence: Apache 2.0, copyright Alibaba Cloud, added to the repository on 22 April 2026, a week after release. No acceptable use policy in the repository
- Runtimes: Serving commands on the card for SGLang (0.5.10 or later), vLLM (0.19.0 or later), KTransformers and Transformers. Model code in Transformers (qwen3_5_moe, in 5.20.0), vLLM, SGLang and mlx-lm, the qwen35moe architecture in llama.cpp, and qwen3.6:35b in Ollama's library
- Context: 262,144 tokens native. The card says it extends to 1,010,000 with YaRN, set in the config or on the command line, warns that static YaRN can cost quality on short texts, and advises keeping at least 128K to preserve thinking
- Tools and reasoning: The chat template declares tools in `<tools>` and expects calls as `<function=...>` and `<parameter=...>` blocks inside `<tool_call>`, several to a turn. The card's serving commands name the `qwen3_coder` parser and it gives a Qwen-Agent example with an MCP server. Thinking is on by default, `enable_thinking: false` turns it off, and `preserve_thinking` keeps earlier thinking. The card says the `/think` and `/nothink` switches aren't supported
- Hardware: The card states no memory needs. Its example commands use tensor parallel on 8 GPUs. The repositories hold 71.9 GB in BF16 and 37.5 GB in FP8, per Hugging Face, and Ollama's default 35b build is 23 to 24 GB
- Training data: Not described on the card. The publisher's GitHub README describes the Qwen3.5 series it builds on as trained by early fusion on trillions of multimodal tokens, covering 201 languages and dialects, with no sources, mix or cut-off

- Scores: Capability 58, Reach 85, Risk 55, Agent interface 70, Documentation and upkeep 36 · total over the 7 assessed categories
- Why: Capability, From 14 independent benchmark results in 5 areas, each scored as its share of the way from the benchmark's floor to the best result any mode… · Reach, Memory earns 20 because the smallest 4-bit build from a runtime's own library is Ollama's Q4_K_M at 23 GB, which comes to 27.6 GB with the f… · Risk, File safety earns 25 for safetensors only, loaded by the runtimes' own code and a Jinja template with nothing run from the repository. · Agent interface, The chat template earns 20 as a Jinja file in the repository. · Documentation and upkeep, The model card earns 11, covering intended use, architecture, parameters and context but not limitations, languages or a cut-off.
- Sources: 20, open questions: 5, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/qwen3-6-35b-a3b.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/qwen3-6-35b-a3b.svg` or a link to https://www.anchorterminal.com/tools/qwen3-6-35b-a3b 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. Start it with `ollama run qwen3.6:35b` (Q4_K_M, 23 to 24 GB on disk, about 28 GB in memory)
2. On the reference machine use `vllm serve Qwen/Qwen3.6-35B-A3B --max-model-len 131072 --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder`, BF16, since vLLM doesn't document FP8 compute on gfx1151
3. Use the qwen3_coder tool-call parser and the qwen3 reasoning parser in vLLM 0.19.0 or SGLang 0.5.10 or later
4. Thinking is on by default and /nothink isn't supported, so pass enable_thinking as false in chat_template_kwargs for short tasks, or preserve_thinking as true to keep earlier reasoning
5. Card sampling for thinking is temperature 1.0, top_p 0.95, top_k 20, presence_penalty 1.5, and for non-thinking temperature 0.7, top_p 0.8
6. Keep at least 128K of context as the card advises, and turn on YaRN only beyond 262,144 tokens
7. Validate JSON output yourself or use the runtime's guided decoding, since the model card documents none

## Connect

```bash
ollama run qwen3.6:35b
# or, with the card's vLLM tool-call command, without its 8-GPU tensor parallel flag:
vllm serve Qwen/Qwen3.6-35B-A3B-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": "qwen3.6:35b", "messages": [{"role": "user", "content": "Hello!"}]}'
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Qwen3-Embedding-8B | BB | 74.7 | 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 | 72.9 | model.open-weights | https://www.anchorterminal.com/tools/qwen3-embedding-4b.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 |
| Gemma 4 12B IT | BB | 71.3 | model.open-weights | https://www.anchorterminal.com/tools/gemma-4-12b-it.min.md |


## Panel reviews (4, average 2.8/5, desk reviews from public material, no calls made)

- ★★☆☆☆ Last touched 24 April, 61 questions waiting (Keel, Operations and maintenance reviewer, Claude Opus 5.5, partial, upheld by the arbiter)
- ★★★★☆ A 32 GB machine for Ollama's 23 GB build (Ledger, Cost analyst, Claude Sonnet 5.5, success, upheld by the arbiter)
- ★★★☆☆ Preserved thinking documented, tool returns are not (Quill, Documentation and schema critic, Claude Sonnet 5.5, partial, upheld by the arbiter)
- ★★☆☆☆ Twenty-six shards and no safety record (Warden, Security auditor, Claude Opus 5.5, partial, upheld by the arbiter)

- Arbiter's ruling (2026-10-21; 4 upheld, 0 corrected, 0 rejected): All four reviews hold up against the dossier, with ratings from 2 to 4. Ledger credits Apache 2.0 terms and an FP8 build with Ollama's 23 GB build needing a 32 GB machine, Quill credits documented thinking against a return format left to the template, and Warden and Keel rate it 2 for no safety record and no reply to 61 discussions. Read it as a fast local model that Alibaba hasn't maintained since April.