Ant Group's inclusionAI releases trillion-parameter reasoning model Ring-2.6-1T
Original title:inclusionAI/Ring-2.6-1T
AI summary
Ant Group's inclusionAI has released the flagship reasoning model Ring-2.6-1T, with a parameter scale of one trillion, targeting complex scenarios such as agent workflows, engineering development, and scientific research analysis. The model upgrades from 'can answer' to 'can execute,' performing better in multi-step tasks and tool collaboration; it supports high/xhigh reasoning intensity levels to flexibly balance performance and cost; and uses asynchronous reinforcement learning combined with the IcePop algorithm to improve training efficiency. In benchmarks, high mode scores 87.60 on PinchBench, 63.82 on ClawEval, and 95.32 on Tau2-Bench in telecom scenarios; xhigh mode scores 66.18 on ARC-AGI-V2, 95.83 on AIME 26, and 88.27 on GPQA Diamond. Context length supports 128K extended to 256K (YaRN), and it is available for download via HuggingFace and ModelScope.
Key Takeaways
- Ant Group's inclusionAI has released the flagship reasoning model Ring-2.
- 6-1T,with a parameter scale of one trillion,targeting complex scenarios such as agent…
- The model upgrades from 'can answer' to 'can execute,' performing better in multi-step tasks and…
- it supports high/xhigh reasoning intensity levels to flexibly balance performance and cost
- and uses asynchronous reinforcement learning combined with the IcePop algorithm to improve…
Quick Facts
- Published
- Updated
- Category
- Model Releases
- AI score
- 59
蚂蚁放出的万亿参数推理模型,Agent执行能力在PinchBench上超GPT-5.4,异步RL训练和可调推理强度对工程落地有参考价值,开源可试。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Ant Group's inclusionAI releases trillion-parameter reasoning model Ring-2.6-1T. May 14, 2026. https://new.nat6.net/en/news/cmp5nc87y0fmvsljx9u9zvsfh/