BAAI releases AREX-2, a 27b agent model
baai has announced arex-2, a 27b-parameter long‑horizon agent model that builds on the qwen3.8 architecture. the model is trained on machine‑learning and algorithmic‑programming tasks with verifiable feedback, and it can iteratively refine solutions through propose‑measure‑reflect‑revise cycles.
Key points
- arex‑2 is a 27b‑parameter agent model based on qwen3.8
- context length is 262,144 tokens
- trained on coding, machine‑learning, and deep‑research tasks
arex‑2 supports a context length of 262,144 tokens and is evaluated on algorithmic programming, machine‑learning engineering, deep research, and general agentic reasoning. the paper reports performance on the frontier‑cs 188‑task agent track and on the mle‑lite any medal benchmark.
the model is released under the apache license 2.0 for research use and can be accessed via the transformers library with the model id "baai/arex‑2".
BAAI/AREX-2 - 27B - Agent model based on Qwen3.8 27B
huggingface.co · 30 September 2026
Loading the full article…
This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Reddit discussionreddit.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Liquid AI launches d1 decision model for structured tasks with zero output tokens · 1 src
- OpenAI scraps GPT-6.1 Astra release over safety concerns · 58 src
- OpenAI hosts DevDay 2026 in San Francisco · 1 src
- AI performance costs have dropped about 47% per quarter, report finds · 2 src
- OpenAI adds GPT-6 Sol and Luna models to ChatGPT and Codex · 47 src
Comments
via GitHub Discussions