--- title: AREX-2 canonical_url: "https://www.modelscope.cn/models/BAAI/AREX-2" md_url: "https://www.modelscope.cn/models/BAAI/AREX-2.md" repository: BAAI/AREX-2 last_updated: 2026-10-13 license: apache-2.0 pipeline_tag: image-text-to-text tasks: - image-text-to-text model_type: - qwen3_5 architectures: - Qwen3_5ForConditionalGeneration base_model: - Qwen/Qwen3.8-27B base_model_relation: finetune parameters: 27.4B tensor_type: - BF16 library_name: - pytorch - transformer - safetensors frameworks: - pytorch downloads: 81 stars: 3 tags: - qwen3_5 - agent - deep-research - reasoning - tool-use - long-context - self-improvement - safetensors --- # AREX-2 > AREX-2 - BAAI 在 ModelScope 开源的模型。AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks BAAI/AREX-2 是 ModelScope 魔搭社区上的 27.4B 参数image-text-to-text模型，采用 apache-2.0 许可，基于 Qwen/Qwen3.8-27B 构建。 - **Repository**: BAAI/AREX-2 - **License**: apache-2.0 - **Tasks**: image-text-to-text - **Parameters**: 27.4B - **Base model**: Qwen/Qwen3.8-27B - **Tags**: qwen3_5, agent, deep-research, reasoning, tool-use, long-context, self-improvement, safetensors - **Downloads**: 81 - **Stars**: 3 - **Last updated**: 2026-10-13 Source: https://www.modelscope.cn/models/BAAI/AREX-2 ---
AREX AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks
Paper Homepage
## Introduction AREX-2 is a 27B-parameter long-horizon agent model from the Beijing Academy of Artificial Intelligence (BAAI). It learns to improve a solution over multiple test-time rounds: propose, measure, reflect, and revise. AREX-2 is trained on machine-learning and algorithmic-programming tasks with verifiable feedback, together with the existing AREX deep-research data. The learned self-improvement behavior transfers to deep research without adding new search trajectories. - **Architecture:** Dense Qwen3.8-compatible multimodal model - **Parameters:** 27B - **Context length:** 262,144 tokens ## Key features - **Long-horizon self-improvement:** turns extra test-time rounds into useful solution refinement. - **Feedback-driven reflection:** reads scores, logs, errors, and timings to decide what to change next. - **Cross-domain performance:** training on coding and machine-learning tasks also improves the model's deep-research performance. - **Long-horizon reasoning:** sustains productive iteration as the task budget grows. ## Evaluation AREX-2 is evaluated on algorithmic programming, machine-learning engineering, deep research, and general agentic reasoning. Results follow the protocols reported in the AREX-2 paper. ### Coding and machine-learning engineering
Model Params Frontier-CS MLE-Lite
Closed-weight models
GPT-5.6 Sol-76.472.7
Claude Opus 4.8-74.563.6
GPT-5.5-72.168.2
Gemini-3.1-Pro-68.9-
Qwen3.7-Max-61.9-
Open-weight models
Kimi-K32.8T-72.7
Naive-N0.5-Flash309B-73.7
DeepSeek-V4-Pro1.6T44.754.5
DeepSeek-V4-Flash284B39.151.5
Kimi-K2.7-Code1T54.7-
GLM-5.3-Flash320B50.4-
Kimi-K2.61T46.966.7
Frontis-MA1-35B35B-71.2
BigBang-V135B-59.1
Qwen3.6-35B-A3B35B23.439.4
AREX-227B70.781.8
Frontier-CS is the 188-task Agent Track. MLE-Lite reports Any Medal averaged over three seeds. ### General agentic reasoning and deep research
Model Params BrowseComp HLE GAIA DeepSearchQA
Frontier models
GPT-5.6 Sol-90.458.0*--
GPT-5.6 Terra-87.5---
GPT-5.6 Luna-83.3---
Kimi-K32.8T91.256.0*-95.0
Claude Fable 5-88.064.5*-94.2
Claude Opus 4.8-84.357.9*-93.1
GPT-5.5-84.452.2*87.4-
Gemini-3.1-Pro-85.951.4*80.693.3
Large models (>40B)
GLM-5744B75.950.470.0-
Kimi-K2.61T83.254.0*80.692.5
GLM-5.3-Flash320B-55.3*78.8-
DeepSeek-V4-Flash284B73.245.157.590.6
DeepSeek-V4-Pro1.6T83.448.271.188.7
MiroThinker-1.7235B74.042.982.772.1
XYZ-Aquila-pro397B84.853.3-92.5
Iris-pro397B88.656.4-92.9
AREX-Base122B82.552.485.489.9
Small models (≤40B)
Tongyi-DeepResearch-30B30B43.432.970.9-
Qwen3.5-35B35B61.047.480.068.5
XYZ-Aquila-mini35B78.851.197.189.5
BigBang-V135B76.550.3--
Quest-35B35B64.637.280.8-
Apodex-1.0-mini35B71.546.8-82.2
Agents-A135B75.547.696.0-
MiroThinker-1.7-mini30B67.936.480.367.9
Iris-mini35B82.252.3-86.9
AREX-Turbo4B70.740.681.678.5
AREX-227B84.052.692.293.8
HLE values marked * are from the full HLE set; unmarked values use the text-only subset. ## Inference Use a recent Transformers release with Qwen3.8 support. ```bash pip install -U torch transformers accelerate ``` ```python import torch from transformers import AutoModelForMultimodalLM, AutoProcessor model_id = "BAAI/AREX-2" processor = AutoProcessor.from_pretrained(model_id) model = AutoModelForMultimodalLM.from_pretrained( model_id, dtype=torch.bfloat16, device_map="auto" ) messages = [{ "role": "user", "content": "Propose a solution and explain how you would improve it over several rounds.", }] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) with torch.inference_mode(): outputs = model.generate(**inputs, max_new_tokens=1024) print(processor.decode( outputs[0][inputs["input_ids"].shape[-1]:], skip_special_tokens=True )) ``` ## Intended use AREX-2 is intended for research on long-horizon agents, iterative problem solving, machine-learning engineering, algorithmic coding, and tool-augmented deep research. ## License AREX-2 is released under the [Apache License 2.0](LICENSE). Follow the terms and notices for the Qwen base model and any downstream data or tools. ## Citation ```bibtex @article{2026arex2, title = {AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks}, author = {Qian, Hongjin and Li, Chaofan and Luo, Kun and Wei, Wenqing and Chen, Jianlyu and Lu, Shuqi and Hu, Yuyang and Xiao, Hongwang and Wang, Hui and Li, Chaozhuo and Ye, Qiwei and Dou, Zhicheng and Lian, Defu and Liu, Zheng}, journal = {arXiv preprint arXiv:2609.38288}, year = {2026}, url = {https://arxiv.org/abs/2609.38288} } ```