88Z AI, 30 April — Ant Baili has officially open-sourced Ling-2.6-1T, a trillion-parameter comprehensive flagship model. Utilizing a hybrid architecture of MLA and Linear Attention, it achieves high intelligence efficiency with minimal token consumption, matching the comprehensive intelligence of GPT-5.4 (Non-Reasoning). The model attains open-source state-of-the-art results on execution benchmarks like AIME26, SWE-bench Verified, and BFCL-V4, demonstrating robust capabilities in agent execution, code generation,
At a glance
- Ling-2.6-1T uses a hybrid architecture for high efficiency, matching GPT-5.4 intelligence
- Achieves open-source SOTA on execution benchmarks like AIME26 with strong agent and coding skills
The update
The Ant Baili large model team has officially open-sourced its latest trillion-parameter comprehensive flagship model, Ling-2.6-1T, designed to handle various complex tasks. The model innovatively adopts a hybrid architecture combining MLA and Linear Attention, achieving a high intelligence-to-efficiency ratio while significantly reducing token consumption. In terms of comprehensive intelligence, Ling-2.6-1T matches the level of GPT-5.4 (Non-Reasoning version). The model excels in multiple key execution benchmarks, achieving state-of-the-art results in the open-source community on lists such as AIME26, SWE-bench Verified, and BFCL-V4. This confirms that Ling-2.6-1T possesses strong practical application capabilities in agent execution, automatic code generation, and long-context understanding, providing developers with a high-performance open-source option.
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Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.