---
title: ExoMind-F16-GGUF
canonical_url: "https://www.modelscope.cn/models/AI4SGI/ExoMind-F16-GGUF"
md_url: "https://www.modelscope.cn/models/AI4SGI/ExoMind-F16-GGUF.md"
repository: AI4SGI/ExoMind-F16-GGUF
last_updated: 2026-09-01
license: apache-2.0
pipeline_tag: image-text-to-text
tasks:
  - image-text-to-text
base_model:
  - AI4SGI/ExoMind
base_model_relation: quantized
library_name:
  - gguf
downloads: 5
stars: 0
tags:
  - exomind
  - gguf
  - llama-cpp
  - f16
  - scientific-reasoning
  - agentic
  - tool-use
  - multimodal
  - qwen3.5
---

# ExoMind-F16-GGUF

> ExoMind-F16-GGUF - AI4SGI 在 ModelScope 开源的模型。ExoMind: Democratizing Scientific Intelligence via Extended-Mind-Inspired Agentic System

AI4SGI/ExoMind-F16-GGUF 是 ModelScope 魔搭社区上的image-text-to-text模型，采用 apache-2.0 许可，基于 AI4SGI/ExoMind 构建。

- **Repository**: AI4SGI/ExoMind-F16-GGUF
- **License**: apache-2.0
- **Tasks**: image-text-to-text
- **Base model**: AI4SGI/ExoMind
- **Tags**: exomind, gguf, llama-cpp, f16, scientific-reasoning, agentic, tool-use, multimodal, qwen3.5
- **Downloads**: 5
- **Stars**: 0
- **Last updated**: 2026-09-01

Source: https://www.modelscope.cn/models/AI4SGI/ExoMind-F16-GGUF

---

<div align="center">

<img src="./assets/ExoMind.png" alt="ExoMind" width="560">

# ExoMind: Democratizing Scientific Intelligence via Extended-Mind-Inspired Agentic System

**ExoMind Team · Shanghai Artificial Intelligence Laboratory**

<p>
  <a href="https://ai4sgi.github.io/ExoMind/"><img src="https://img.shields.io/badge/Project_Page-Visit-174F87?style=for-the-badge&logo=googlechrome&logoColor=white" alt="Project Page"></a>
  <a href="https://doi.org/10.20944/preprints202608.2038.v1"><img src="https://img.shields.io/badge/Paper-Preprint-B31B1B?style=for-the-badge&logo=adobeacrobatreader&logoColor=white" alt="ExoMind preprint"></a>
</p>
<p>
  <a href="https://huggingface.co/AI4SGI/ExoMind-F16-GGUF"><img src="https://img.shields.io/badge/Hugging_Face-Model-FFD21E?style=for-the-badge&logo=huggingface&logoColor=000000" alt="Hugging Face"></a>
  <a href="https://github.com/AI4SGI/ExoMind"><img src="https://img.shields.io/badge/GitHub-Code-181717?style=for-the-badge&logo=github&logoColor=white" alt="GitHub"></a>
  <a href="https://modelscope.cn/models/AI4SGI/ExoMind-F16-GGUF"><img src="https://img.shields.io/badge/ModelScope-Model-624AFF?style=for-the-badge" alt="ModelScope"></a>
</p>

</div>

## Overview

Reference-precision GGUF release of [ExoMind](https://huggingface.co/AI4SGI/ExoMind). Download all four numbered shards and pass the first shard to llama.cpp; the remaining shards are discovered automatically.

This repository intentionally contains only the **F16** model and
the matching multimodal projector. Keeping each precision in its own repository
makes downloads, local disk requirements, and deployment commands explicit.

## Files

| File | Role | Download size |
| --- | --- | ---: |
| `qwen3_5_35b_a3b-F16-00001-of-00004.gguf` | F16 shard 1/4 | 18.15 GiB |
| `qwen3_5_35b_a3b-F16-00002-of-00004.gguf` | F16 shard 2/4 | 18.31 GiB |
| `qwen3_5_35b_a3b-F16-00003-of-00004.gguf` | F16 shard 3/4 | 18.18 GiB |
| `qwen3_5_35b_a3b-F16-00004-of-00004.gguf` | F16 shard 4/4 | 11.55 GiB |
| `mmproj-qwen3_5_35b_a3b-F16.gguf` | F16 multimodal projector | 857.62 MiB |

## Quick Start with llama.cpp

Text-only serving:

```bash
llama-server \
  -m qwen3_5_35b_a3b-F16-00001-of-00004.gguf \
  --ctx-size 32768 \
  --host 0.0.0.0 \
  --port 8080
```

For image input, load the projector shipped in this repository:

```bash
llama-server \
  -m qwen3_5_35b_a3b-F16-00001-of-00004.gguf \
  --mmproj mmproj-qwen3_5_35b_a3b-F16.gguf \
  --ctx-size 32768 \
  --host 0.0.0.0 \
  --port 8080
```

## Conversion Provenance

The original 71,066,994,432-byte F16 GGUF was split with `llama-gguf-split` at a 20 GB maximum shard size using [llama.cpp revision `7584430`](https://github.com/ggml-org/llama.cpp/commit/7584430716ee229751771ed0d6bbcb780d105eeb). The four shards contain all 753 tensors, add 128 bytes of split metadata, and passed the pinned tool's merge dry run. The original unsplit file remains intact. These GGUF files were supplied as existing release artifacts. Their exact filenames, byte sizes, and GGUF v3 headers were validated before publication, but the original HF-to-GGUF conversion and quantization commands were not retained with the files. The repository therefore does not claim bit-for-bit reproducibility of the original conversion pipeline.

## Evaluation Boundary

Published benchmark results use the original Transformers BF16 checkpoint. The F16 GGUF packaging has not been assigned separate scores.

Complete settings and comparisons are available in the
[evaluation explorer](https://ai4sgi.github.io/ExoMind/#results).

## License and Attribution

The model files and upstream Qwen3.5 materials are distributed under the Apache
License 2.0 included with the model. Preprint text, scientific figures,
results, and ExoMind brand assets are governed by the
[ExoMind Research Content and Brand Terms](./CONTENT_RIGHTS.md). See
[NOTICE.md](./NOTICE.md) for third-party notices.

## Citation

```bibtex
@article{Ye_2026,
  title     = {ExoMind: Democratizing Scientific Intelligence via Extended-Mind-Inspired Agentic System},
  author    = {Ye, Peng and Liu, Zhuo and Ye, Jingqi and Yu, Fangchen and Tang, Shengji and Jiang, Yichen and He, Haonan and Cao, Zongsheng and Chen, Tao and Zhang, Bo and Ouyang, Wanli and Zhou, Bowen and Bai, Lei},
  year      = {2026},
  month     = aug,
  publisher = {MDPI AG},
  doi       = {10.20944/preprints202608.2038.v1},
  url       = {https://doi.org/10.20944/preprints202608.2038.v1}
}
```