Ring-2.6-1T: Trillion-Parameter Model for Agent Workflows and Reasoning

LinkinclusionAIResource

inclusionAI has released Ring-2.6-1T, a trillion-parameter reasoning model under the MIT licence. It is designed for multi-step agent tasks, offering adjustable reasoning intensity and asynchronous RL training.

Key points

  1. Trillion-parameter open-weight model released under the MIT licence on Hugging Face and ModelScope.
  2. Context window supports 128K extended to 256K tokens via YaRN.
  3. Features two reasoning effort levels: 'high' for standard agent workflows and 'xhigh' for complex math and research tasks.
  4. Trained using an asynchronous RL (Async RL) architecture and the IcePop algorithm to ensure stability at trillion-parameter scale.
  5. Benchmark results include 87.60 on PinchBench ('high') and 66.18 on ARC-AGI-V2 ('xhigh').
  6. Ready for multi-node deployment with SGLang and vLLM, supporting both BF16 and FP8 precisions.

Why it matters

Trillion-parameter open models are closing the capability gap with frontier proprietary systems in agent execution and deep reasoning. Configurable reasoning tiers enable developers to balance latency and computational depth in production workflows.

Analysis

inclusionAI tackles training bottlenecks and idle GPU resources by decoupling policy sampling from parameter updates via Async RL. Together with the IcePop algorithm, this aims to avoid policy degradation over long-horizon runs. By focusing on multi-turn tool collaboration, task planning, and error recovery, the release targets production-grade coding and enterprise automation agents.

Mentioned links

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Link:Kilo Code

    Ring-2.6-1T: Free inclusionAI Thinking Model on Kilo Code

    inclusionAI has released Ring-2.6-1T, a 1T-parameter-scale thinking model featuring 63 billion active parameters. Optimised for coding agents, the model is accessible free of charge via Kilo Code and OpenRouter.

    AI & LLMs· Tool

  • Link:inclusionAI

    Ling-3.0-tiny: Lightweight Hybrid Reasoning MoE Model by inclusionAI

    inclusionAI has introduced Ling-3.0-tiny, a compact MoE model with 7.9B total parameters activating only 1.3B parameters per token. Utilizing a hybrid architecture of linear and latent attention mechanisms, it aims to deliver efficient reasoning and agentic workloads on local hardware.

    AI & LLMs· Tool

  • Link:Ant Group

    inclusionAI: Ant Group's AI Initiative on Hugging Face

    inclusionAI serves as the Hugging Face hub for Ant Group's Artificial General Intelligence (AGI) initiative, hosting open-source LLMs, image generation models, and reinforcement learning systems.

    AI & LLMs· Resource

  • Repository:ornith-ai/Ornith-1

    Ornith: Self-Improving Open-Source Models for Agentic Workflows

    The Ornith-1 repository provides an open-source model family designed for agentic tasks using autonomous self-improvement. Spanning Ornith-1.0 and Ornith-1.5, the project implements self-scaffolding and automated generation of training rollouts for reinforcement learning.

    2kstars

    AI & LLMs· Tool

  • Link:ornith-ai

    Ornith-1.5-397B: Open-Source MoE Model Focused on Coding and Reasoning

    ornith-ai has released Ornith-1.5-397B, an open-source multimodal mixture-of-experts model under the MIT licence. With 397 billion parameters and an end-to-end self-improvement loop, it delivers coding and reasoning benchmark performance comparable to top proprietary models.

    AI & LLMs· Resource

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.