Back to AI News
AITopic existing content

Ant Group Open-Sources Ling-2.6-1T: A Trillion-Parameter MoE Model with Fast-Thinking Capability

Ant Group&#39;s Inclusion AI has officially released the weights for its flagship large language model,&nbsp;Ling-2.6-1T. This trillion-parameter MoE (Mixture of Experts) model is now available to the public under a permissive MIT license. While feat<!--autointro-->

Ant Group's Inclusion AI has officially released the weights for its flagship large language model, Ling-2.6-1T. This trillion-parameter MoE (Mixture of Experts) model is now available to the public under a permissive MIT license. While featuring a total parameter count of 1 trillion, it activates only 63 billion parameters during inference, balancing immense capability with computational efficiency. The model supports a context window of up to 256K tokens.

Architecture and Core Features

Ling-2.6-1T shares the same advanced architecture as the recently open-sourced lightweight "flash" version (104B total / 7.4B active parameters). It employs a hybrid attention mechanism combining MLA (Multi-head Latent Attention) and Lightning Linear attention.
A key innovation in the 1T version is the introduction of a "Fast-Thinking" training strategy. By incorporating a "Contextual Process Redundancy Suppression" reward during post-training, the model is optimized to compress verbose chain-of-thought outputs. This approach significantly reduces token consumption without compromising the quality of reasoning or task execution.

Performance Benchmarks

The model demonstrates state-of-the-art performance among open-source non-reasoning models on several complex task benchmarks:
  • SWE-bench Verified: Achieved a score of 72.2%, a notable improvement over the flash version's 61.2%. For context, some open-source reasoning models like DeepSeek V4 Pro have surpassed 80%.

  • Other Benchmarks: The model reportedly reaches top-tier performance on execution-focused tests including AIME 2026, BFCL-V4, TAU2-Bench, and IFBench.

  • Efficiency: On the Artificial Analysis comprehensive intelligence index, the model scored 34, consuming approximately 16 million output tokens for the full evaluation, highlighting its token efficiency.

Deployment and Accessibility

Ling-2.6-1T is designed for integration into complex workflows and is compatible with mainstream agent frameworks such as Claude Code, OpenClaw, and OpenCode.


Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!