---
title: QwenPaw-Flash-27B-BF16
canonical_url: "https://www.modelscope.cn/models/AgentScope/QwenPaw-Flash-27B-BF16"
md_url: "https://www.modelscope.cn/models/AgentScope/QwenPaw-Flash-27B-BF16.md"
repository: AgentScope/QwenPaw-Flash-27B-BF16
last_updated: 2026-11-02
license: apache-2.0
pipeline_tag: text-generation
tasks:
  - text-generation
library_name:
  - gguf
downloads: 2
stars: 0
tags:
  - qwenpaw
  - tool-use
  - gguf
---

# QwenPaw-Flash-27B-BF16

> QwenPaw-Flash-27B-BF16 - AgentScope 在 ModelScope 开源的模型。Hugging Face mirror preparation; weights pending synchronization confirmation.

AgentScope/QwenPaw-Flash-27B-BF16 是 ModelScope 魔搭社区上的text-generation模型，采用 apache-2.0 许可。

- **Repository**: AgentScope/QwenPaw-Flash-27B-BF16
- **License**: apache-2.0
- **Tasks**: text-generation
- **Tags**: qwenpaw, tool-use, gguf
- **Downloads**: 2
- **Stars**: 0
- **Last updated**: 2026-11-02

Source: https://www.modelscope.cn/models/AgentScope/QwenPaw-Flash-27B-BF16

---

# QwenPaw-Flash-27B-BF16

QwenPaw-Flash-27B-BF16 is the BF16 GGUF release of the QwenPaw-Flash-27B agent model. The QwenPaw-Flash family is trained for tool use, command execution, memory management and multi-step tasks in QwenPaw.

[Project](https://github.com/agentscope-ai/QwenPaw) · [Hugging Face](https://huggingface.co/agentscope-ai/QwenPaw-Flash-27B-BF16) · [ModelScope collection](https://www.modelscope.cn/collections/AgentScope/QwenPaw-Flash)

## Download from ModelScope

With the ModelScope CLI installed, download this repository into one directory.

```bash
modelscope download --model AgentScope/QwenPaw-Flash-27B-BF16 --local_dir QwenPaw-Flash-27B-BF16
```

## Weight files

Download all numbered shards into the same directory. Pass the first shard to llama.cpp.

| File | Bytes | SHA-256 |
| --- | ---: | --- |
| `QwenPaw-Flash-27B-step92-BF16-00001-of-00002.gguf` | 29,955,252,352 | `656d7d75fef1c014f21bc3144530b5c0e94f8af719cf872131b13950ae2f882d` |
| `QwenPaw-Flash-27B-step92-BF16-00002-of-00002.gguf` | 23,853,028,672 | `f75d6f1628f40bd9d4fd33a2497e726074ba178b915050b1e020786cefaf8e93` |

The SHA-256 of the original GGUF before splitting was `35d9db421d316780b9cefc9956a6250742c20457ce7965f1bea8abe4fd3334e7`.

## Serve locally

Use a llama.cpp build that supports this architecture.

```bash
llama-server -m QwenPaw-Flash-27B-BF16/QwenPaw-Flash-27B-step92-BF16-00001-of-00002.gguf --alias QwenPaw-Flash-27B-BF16 --port 8080
```

Send a text request to the running server.

```bash
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model":"QwenPaw-Flash-27B-BF16","messages":[{"role":"user","content":"Hello, QwenPaw!"}],"max_tokens":128}'
```

## License

Apache License 2.0.
