# Qwen-AgentWorld-35B-A3B (slim)

> Qwen-AgentWorld-35B-A3B is Alibaba's open-weight language world model, a 35B mixture of experts with 3B active that simulates agent environments by predicting the next observation across seven domains, published in BF16 under Apache 2.0.

- Full: https://www.anchorterminal.com/tools/qwen-agentworld-35b-a3b.md (~10,500 tokens) · this version ~1,680 tokens · JSON https://www.anchorterminal.com/tools/qwen-agentworld-35b-a3b.json · canonical https://www.anchorterminal.com/tools/qwen-agentworld-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: 2026-11-10 · pricing https://www.anchorterminal.com/tools/qwen-agentworld-35b-a3b#pricing · reviews https://www.anchorterminal.com/tools/qwen-agentworld-35b-a3b#reviews
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)

**D · 52.1/100 · rank #122 of 135 open-weight models · #94 in Open-weight models · not agent-ready · confidence low**


Assessment: A specialist, not a chat model. It plays the environment an agent acts in, across seven domains, so you can train or test an agent without the real thing. Alibaba ships BF16 only, which fits a 128 GB machine, and hasn't answered a discussion on it or changed the files since June.

## 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 66/100
- Packages: huggingface `Qwen/Qwen-AgentWorld-35B-A3B`
- 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/Qwen/Qwen-AgentWorld-35B-A3B, read from the model card, the files and their metadata, the chat template and the runtimes' own code at their latest releases on 10 November 2026. The technical report, the blog and the GitHub repository weren't read. Nothing was downloaded or run
- Model: 34,660,610,688 parameters in BF16 per the safetensors metadata, the Qwen3.5 mixture-of-experts architecture (Qwen3_5MoeForConditionalGeneration), 40 layers, 256 experts with 8 routed and 1 shared per token, about 3B active. A language world model, trained to predict the next environment state from an agent's action and the history. Text only in this checkpoint
- Files: 21 safetensors shards in BF16 (69.3 GB), a chat_template.jinja that matches the template in tokenizer_config.json, 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 in the repository. Nothing is added on the card
- Runtimes: The architecture is Qwen3.5's mixture of experts, which released vLLM (0.33.0), SGLang (0.5.23), Transformers (5.22.2) and llama.cpp (b12242) run. The card gives SGLang, vLLM and Transformers commands. Ollama's library has no entry for it, and Alibaba publishes no GGUF
- Context: 262,144 tokens, per the card and max_position_embeddings in the text config. The card advises keeping at least 128K, since the simulation leans on long histories. tokenizer_config.json says 131,072
- Tools and reasoning: It thinks by default before giving the predicted observation, and the card names the qwen3 reasoning parser for vLLM and SGLang. The template is Qwen3.5's, with tool declarations and tool_call blocks, but the card presents the model as a simulator of tool environments and doesn't describe calling tools with it
- Hardware: 69.3 GB in BF16, about 83 GB with the fifth added, which fits a 128 GB unified machine. The card's commands use four GPUs. Unsloth's UD-Q4_K_M GGUF is 22.1 GB and its Q8_0 36.9 GB
- Training data: The card names the three stages and says no outputs from external API services were included. No datasets, sizes or cut-off on the card

- Scores: Capability pending, Reach 55, Risk 55, Agent interface 50, Documentation and upkeep 41 · total over the 7 assessed categories
- Why:  · Reach, Memory 10, Alibaba's only build is the BF16 safetensors, 69.32 GB, and 69.32 GB × 1.2 = 83.2 GB, above 64 GB and inside the reference machin… · Risk, File safety 25, 21 safetensors shards and no pickle or Python files, and config.json has no auto_map. · Agent interface, Chat template 20, chat_template.jinja in the repository, the same as the template in tokenizer_config.json. · Documentation and upkeep, Model card 11, the card gives intended use (simulating agent environments), architecture, parameters and context length.
- Sources: 15, open questions: 6, both in the full twin
- Capabilities: model.open-weights
- JSON: https://www.anchorterminal.com/api/v1/tools/qwen-agentworld-35b-a3b.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/CV4E9rpNSD; https://github.com/QwenLM/Qwen3.8/issues; https://huggingface.co/Qwen/Qwen-AgentWorld-35B-A3B/discussions; more in the full twin

## Before you call it

1. Use it as the environment, not the agent. Give it the domain's system prompt from Alibaba's GitHub repository, the history and the agent's action, and it returns the predicted observation
2. With vLLM pass `--language-model-only`, or startup fails on the missing vision weights
3. Use temperature 0.6, top_p 0.95 and top_k 20, and allow up to 32,768 output tokens, as the card recommends
4. Keep at least 128K of context, since the simulation depends on the history
5. Thinking is on by default and comes back through the qwen3 reasoning parser
6. A simulated observation is a prediction. Check anything that matters against the real environment

## Connect

```bash
# vLLM, as the card gives it (four GPUs)
vllm serve Qwen/Qwen-AgentWorld-35B-A3B --port 8000 --tensor-parallel-size 4 \
  --max-model-len 262144 --reasoning-parser qwen3 --language-model-only --trust-remote-code
# or, with llama.cpp and Unsloth's GGUF, a third party's conversion:
llama-server -hf unsloth/Qwen-AgentWorld-35B-A3B-GGUF:UD-Q4_K_M --jinja
```

```bash
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "Qwen/Qwen-AgentWorld-35B-A3B", "messages": [{"role": "user", "content": "Hello!"}], "temperature": 0.6}'
```

## 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)