What license applies to Arx12/AREX-Base-GGUF?

License: apache-2.0. Verify terms on Hugging Face before commercial use.

How do I run Arx12/AREX-Base-GGUF locally?

Download a GGUF file from this page and load it in guIDE or llama.cpp. Pipeline task: text-generation.

Model Intelligence Sheet

Arx12/AREX-Base-GGUF overview

AREX Base GGUF Introduction AREX is a family of deep research agents developed by the Beijing Academy of Artificial Intelligence BAAI . It is designed for long…

llama.cppggufagentdeep-researchreasoningtool-uselong-contextqwen3.5mixture-of-expertstext-generationenbase_model:Qwen/Qwen3.5-122B-A10Bbase_model:quantized:Qwen/Qwen3.5-122B-A10Blicense:apache-2.0endpoints_compatibleregion:usconversational

Runs locally from ~9.11 GB disk (12 GB VRAM class GPUs with llama.cpp / guIDE).

Downloads

Likes

Pipeline

text-generation

Author

Arx12

Repository Files & Downloads

8 GGUF files detected

Direct downloads for local inference

File	Type	Quantization	Size	Link
BF16/AREX-Base-BF16-00001-of-00006.gguf	GGUF	BF16	43.80 GB	Download
BF16/AREX-Base-BF16-00002-of-00006.gguf	GGUF	BF16	43.56 GB	Download
BF16/AREX-Base-BF16-00003-of-00006.gguf	GGUF	BF16	43.63 GB	Download
BF16/AREX-Base-BF16-00004-of-00006.gguf	GGUF	BF16	43.80 GB	Download
BF16/AREX-Base-BF16-00005-of-00006.gguf	GGUF	BF16	43.63 GB	Download
BF16/AREX-Base-BF16-00006-of-00006.gguf	GGUF	BF16	9.11 GB	Download
Q4_K_M/AREX-Base-Q4_K_M.gguf	GGUF	Q4_K_M	69.12 GB	Download
Q8_0/AREX-Base-Q8_0.gguf	GGUF	Q8_0	120.95 GB	Download

Model Details

Model ID	Arx12/AREX-Base-GGUF
Author	Arx12
Pipeline	text-generation
License	apache-2.0
Base model	Qwen/Qwen3.5-122B-A10B
Last modified	2026-07-27T11:13:14.000Z

Model README

---

license: apache-2.0

pipeline_tag: text-generation

library_name: llama.cpp

base_model:

Qwen/Qwen3.5-122B-A10B

language:

tags:

agent
deep-research
reasoning
tool-use
long-context
qwen3.5
mixture-of-experts

---

AREX-Base-GGUF

Introduction

AREX is a family of deep research agents developed by the Beijing Academy of Artificial Intelligence (BAAI). It is designed for long-horizon tasks in which an agent must search across sources, assemble candidate answers, verify multiple constraints, and revise its research plan when the available evidence is incomplete.

AREX follows a recursive self-improvement framework consisting of an inner research loop and an outer self-improvement loop. The inner loop conducts deep research, while the outer loop evaluates and improves the research process.

This repository contains AREX-Base, the larger model in the AREX family. It is built on Qwen3.5-122B-A10B and uses a Mixture-of-Experts architecture with 122B total parameters and 10B activated parameters.

Architecture: Mixture-of-Experts
Parameters: 122B
Context length: 262,144 tokens

Key Features

Recursive self-improvement: evaluates provisional answers and chooses whether to accept, refine, or restart the research trajectory.
Verification-guided research: converts unresolved answer constraints into focused follow-up searches.
Autonomous context updates: retains verified findings, source identifiers, rejected candidates, open constraints, and the next research plan.
Long-horizon tool use: supports multi-round search, browsing, evidence integration, and answer construction.
Evidence-grounded answers: maintains supporting evidence and answer-level confidence throughout the research process.

Model Family

| --- | --- | ---: | --- |

Method

AREX organizes deep research as two interacting loops supported by autonomous context management.

Inner research loop. The agent searches, reads, integrates evidence, tracks candidates, and produces a provisional answer with supporting evidence and an answer-level confidence score.
Outer self-improvement loop. The agent checks the provisional answer against the original constraints. High-confidence answers are accepted. Recoverable trajectories are refined around unresolved claims, while uninformative trajectories can be restarted.
Autonomous context update. The model invokes update_context to refresh its research state around verified findings, current and rejected candidates, unresolved constraints, source validity, and the next research plan.

This design preserves useful work across iterations and reduces repeated exploration of paths that have already been rejected.

Evaluation

AREX is evaluated through a unified long-horizon search-agent interface with search, visit, update_context, and finish tools. HLE with tools additionally provides a Python tool.

<th style="padding:10px 10px;text-align:left;font-weight:600;border-bottom:2px solid #0F766E;color:#0F766E">Model</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">Total<br>Params</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">BrowseComp</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">xbench-<br>2510</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">DeepSearch<br>QA</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">WideSearch-<br>en</th>

<th style="padding:10px 8px;text-align:center;font-weight:500;border-bottom:2px solid #0F766E;color:#0F766E;font-size:14px">HLE w/<br>tools</th>

</tr></thead>

<tbody>

<tr><td colspan="8" style="padding:8px 12px;font-weight:600;color:#0F766E;border-bottom:1px solid rgba(15,118,110,0.2);background:rgba(15,118,110,0.1)">Frontier models</td></tr>

<tr>

</tr>

<tr>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">Gemini-3.1-Pro</td>

</tr>

<tr><td colspan="8" style="padding:8px 12px;font-weight:600;color:#0F766E;border-bottom:1px solid rgba(15,118,110,0.2);background:rgba(15,118,110,0.1)">Open-source models</td></tr>

<tr>

</tr>

<tr>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">DeepSeek-V4-Flash</td>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">DeepSeek-V4-Pro</td>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">Tongyi-DeepResearch</td>

</tr>

<tr>

</tr>

<tr>

</tr>

<tr>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">MiroThinker-1.7-mini</td>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">MiroThinker-1.7</td>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">MiroThinker-H1</td>

</tr>

<tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15)">Quest-35B</td>

</tr>

<tr><td colspan="8" style="padding:8px 12px;font-weight:600;color:#0F766E;border-bottom:1px solid rgba(15,118,110,0.2);background:rgba(15,118,110,0.1)">AREX models</td></tr>

<td style="padding:7px 10px;padding-left:20px;border-bottom:1px solid rgba(128,128,128,0.15);color:#0F766E">AREX-Turbo</td>

</tr>

</tr>

</tbody>

</table>

<div style="font-size:12px;color:#6b7280;margin-top:6px;margin-bottom:-8px;line-height:1.5">* Results reported on the full HLE. Unmarked HLE results use the text-only subset.</div>

</div>

Inference

See inference/README.md for OpenAI-compatible serving and generation examples. The folder also contains the complete, directly formattable system and user prompts used for BrowseComp.

Intended Use

AREX-Base is intended for research-agent applications that require long-horizon information seeking, evidence aggregation, multi-constraint verification, and tool-augmented reasoning. Its larger capacity is suited to settings where research quality is prioritized over serving cost.

License

AREX-Base is released under the Apache License 2.0.

Run Arx12/AREX-Base-GGUF with guIDE

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.

Download guIDE → · Browse 524k+ models · Compare models

Source: Hugging Face · Compare models