Ant Group open-sources compact MoE model Ling-3.0-tiny

ArticleZhidxAnnouncement

Ant Group's Bailian team has released Ling-3.0-tiny, an open-source lightweight mixture-of-experts model. It leverages DeepSeek's MLA and Kimi's KDA architectures with 7.9B total parameters, activating just 1.3B parameters per token for local edge deployment on hardware like Apple Silicon Macs and DGX systems.

Key points

  1. Ling-3.0-tiny features 7.9B total parameters but activates only 1.3B per token across 8 routing experts and 1 shared expert out of 128.
  2. The architecture stacks Moonshot AI's KDA (Kimi Dilation Attention) and DeepSeek's MLA (Multi-head Latent Attention) in an alternating 3:1 ratio.
  3. Native hybrid inference enables rapid responses or extended reasoning per request using the enable_thinking parameter.
  4. Scored 25 on the Artificial Analysis Intelligence Index and 16 on the Agentic Index, outperforming models such as Gemma-4-31B in agentic benchmarks.
  5. Released in BF16, FP8, and INT4 formats, reaching 86–90 tokens/s on an M4 Pro MacBook with 8.34 GiB peak memory at an 8K context.

Why it matters

Running capable agentic workflows locally without cloud API dependencies solves major latency, cost, and data privacy challenges. A model that activates only 1.3B parameters while matching mid-sized models in task execution makes local AI deployment viable on consumer workstations.

Analysis

Ling-3.0-tiny highlights the practical integration of recent Chinese architecture advances, merging DeepSeek's MLA with Moonshot AI's KDA. By minimising activated parameters without sacrificing agent reasoning benchmarks, Ant Group focuses directly on edge and desktop use cases. Future development will target factual accuracy and complex tool-use stability.

Mentioned links

Found in

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: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

  • Repository:inclusionAI/Ling

    Ling: Open-Source Mixture-of-Experts LLMs by InclusionAI

    InclusionAI has open-sourced Ling, a family of Mixture-of-Experts (MoE) large language models. The suite is offered in two sizes: Ling-lite with 16.8 billion total parameters (2.75 billion activated) and Ling-plus with 290 billion total parameters (28.8 billion activated).

    267starsPython

    AI & LLMs· Resource

  • Article:Chris Alexiuk and Chintan Patel

    NVIDIA Nemotron 3.5 Lightning: Compact MoE Model for Agentic Execution

    NVIDIA has announced Nemotron 3.5 Lightning, an open 30B mixture-of-experts (MoE) model featuring 3B active parameters. It is tailored for the high-volume execution layer of long-running AI agents, targeting routine tasks such as tool calling and result validation with low latency.

    AI & LLMs· Announcement

  • 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

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.