# Ant Ling — In RL we trust — AReaL v0.2 (Boba) Release

- Company: Ant Ling (ant-ling.com)
- Announced: 2025-04-01
- Category: open-weight-release
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://developer.ant-ling.com/en/blogs/areal-boba
- Record: https://forck.live/items/22345-in-rl-we-trust-areal-v0-2-boba-release
- Subject: Ling / Ring / Ming
- Open weights: yes
- Models affected: AReaL-boba-RL-7B, AReaL-boba-SFT-32B

Ant Ling released AReaL v0.2 (Boba), an open-source reinforcement learning framework. The release includes a SOTA 7B model for mathematical reasoning (pass@1 61.9 on AIME24, 48.3 on AIME25) and a competitive 32B model trained with only 200 data samples that matches QwQ-32B on AIME 2024. The framework adds SGLang support and achieves 1.5x speed improvement over v0.1 on 7B models.

## Evidence

Verbatim from https://developer.ant-ling.com/en/blogs/areal-boba:

> We obtain a SOTA 7B model in mathematical reasoning, achieving pass@1 score of 61.9 on AIME24 and 48.3 on AIME25 respectively. The highly competitive 32B model was trained with extremely low cost, achieving results comparable to QwQ-32B using only 200 data samples.

---

Record: https://forck.live/items/22345-in-rl-we-trust-areal-v0-2-boba-release
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md