---
title: Piccolo-27B-BF16
canonical_url: "https://www.modelscope.cn/models/AgentScope/Piccolo-27B-BF16"
md_url: "https://www.modelscope.cn/models/AgentScope/Piccolo-27B-BF16.md"
repository: AgentScope/Piccolo-27B-BF16
last_updated: 2026-11-02
license: apache-2.0
pipeline_tag: text-classification
tasks:
  - text-classification
base_model:
  - agentscope-ai/Piccolo-27B
base_model_relation: quantized
library_name:
  - gguf
language:
  - en
  - zh
downloads: 12
stars: 0
tags:
  - gguf
  - bf16
  - decision-model
  - bidirectional-attention
  - qwenpaw
---

# Piccolo-27B-BF16

> Piccolo-27B-BF16 - AgentScope 在 ModelScope 开源的模型。Mirror of agentscope-ai/Piccolo-27B-BF16

AgentScope/Piccolo-27B-BF16 是 ModelScope 魔搭社区上的text-classification模型，采用 apache-2.0 许可，基于 agentscope-ai/Piccolo-27B 构建。

- **Repository**: AgentScope/Piccolo-27B-BF16
- **License**: apache-2.0
- **Tasks**: text-classification
- **Base model**: agentscope-ai/Piccolo-27B
- **Tags**: gguf, bf16, decision-model, bidirectional-attention, qwenpaw
- **Downloads**: 12
- **Stars**: 0
- **Last updated**: 2026-11-02

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

---

# Piccolo-27B-BF16

<div style="display:flex;max-width:calc(100vw - 7rem);flex-wrap:wrap;align-items:center;gap:6px;margin:12px 0">
<a style="display:inline-flex;margin:0" href="https://github.com/agentscope-ai/Piccolo"><img style="display:block;margin:0" src="https://www.modelscope.cn/models/AgentScope/Piccolo-27B/resolve/master/assets/links/project.png" width="87" height="28" alt="Project"></a>
<a style="display:inline-flex;margin:0" href="https://huggingface.co/agentscope-ai/Piccolo-27B-BF16"><img style="display:block;margin:0" src="https://www.modelscope.cn/models/AgentScope/Piccolo-27B/resolve/master/assets/links/huggingface.png" width="128" height="28" alt="Hugging Face"></a>
<a style="display:inline-flex;margin:0" href="https://www.modelscope.cn/collections/AgentScope/QwenPaw-Flash"><img style="display:block;margin:0" src="https://www.modelscope.cn/models/AgentScope/Piccolo-27B/resolve/master/assets/links/modelscope.png" width="183" height="28" alt="ModelScope collection"></a>
<a style="display:inline-flex;margin:0" href="https://decisionplayground.agentscope.io/"><img style="display:block;margin:0" src="https://www.modelscope.cn/models/AgentScope/Piccolo-27B/resolve/master/assets/links/decision.png" width="169" height="28" alt="Decision Playground"></a>
</div>

Piccolo-27B-BF16 is the BF16 GGUF release of Piccolo-27B for local tool review and context selection.

## Download

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

## Precision and conversion

The text backbone was converted directly from the full checkpoint to BF16 GGUF. Download all 12 shards together and load `Piccolo-27B-BF16-00001-of-00012.gguf`.

## Required decision inference

Use the [Piccolo-compatible llama.cpp runtime](https://github.com/agentscope-ai/Piccolo/blob/main/docs/QWENPAW.md) with bidirectional prompt attention and native candidate scoring. The `/v1/systemone` endpoint returns probabilities without generating answer tokens.

## Numerical evidence and limits

On four test questions, the largest candidate-probability difference from the full checkpoint was 0.001725. [Validation details](https://www.modelscope.cn/models/AgentScope/Piccolo-27B-BF16/resolve/master/validation.json) document the comparison.

## License

[Apache 2.0](https://www.modelscope.cn/models/AgentScope/Piccolo-27B-BF16/resolve/master/LICENSE). See [third-party notices](https://www.modelscope.cn/models/AgentScope/Piccolo-27B-BF16/resolve/master/THIRD_PARTY_LICENSES.md).