88Z AI, 9 May — Ant Bailin releases the trillion-parameter model Ring-2.6-1T, significantly enhancing reasoning and multimodal capabilities while reducing inference costs. The API is now open for various task scenarios.
At a glance
- Ring-2.6-1T features a trillion parameters, significantly boosting reasoning and multimodal capabilities while lowering inference costs.
- The model API is now open, supporting text, image, and code tasks, thereby enriching the domestic large model ecosystem.

The update
Ant Bailin has officially launched its flagship reasoning model, Ring-2.6-1T, featuring a trillion-parameter scale. The model achieves significant breakthroughs in core capabilities such as reasoning, programming, and multimodal understanding, leading to comprehensive performance improvements. By employing innovative training architecture optimization techniques, the Bailin team has effectively reduced inference costs while ensuring high-performance output. Currently, Ring-2.6-1T is available via API for developers and enterprise users, fully supporting diverse task scenarios including text generation, image processing, and code writing. This launch further enriches the choices within the domestic large model ecosystem, providing stronger technical support for industry applications.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 百灵大模型心Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.