Ant Ling (Bailing)
Ant Group's open-weight Ling and Ring model family, including trillion-parameter models released free on Hugging Face and ModelScope.
Overview
Ling (also called Bailing) is an open-weight large language model family from Ant Group's Inclusion AI team. It spans the Ling non-thinking series, including trillion-parameter models such as Ling-1T and the upgraded Ling-2.6-1T, and the Ring reasoning series. The models use a mixture-of-experts design that activates only a fraction of parameters per token for efficiency, and are released free under open weights on Hugging Face and ModelScope.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 24 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
ASEAN Perspective
Ant Ling (Bailing) in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
What this is for: An open-weight model family covering fast-response Ling models (up to a trillion parameters) and Ring reasoning models, built on efficient mixture-of-experts architectures.
Who this is for: Developers and researchers who want to self-host large open-weight Chinese-origin models, including trillion-parameter flagships, without licensing fees.
Availability: Free, open weights on Hugging Face and ModelScope. Developed by Ant Group's Inclusion AI (Bailing) team.
What people say
Ant Group's Inclusion AI team earned strong open-source credibility with the Ling and Ring families, and the benchmark reception has been unusually warm for a Chinese-origin release. DeepLearning.AI's The Batch highlighted Ling-1T as an open non-reasoning model that outperforms closed competitors, while the Ring reasoning series posted headline results: around 70.42% on AIME 2025, a #1 open-source finish on ArtifactsBench, roughly 81.6% on Arena-Hard V2 (approaching GPT-5-Thinking), and an IMO silver-medal-equivalent showing billed as a first for an open system. The MIT license on the trillion-parameter weights, published to Hugging Face and ModelScope, is a genuine draw for teams that want no licensing friction.
The obvious limitation is practicality: these are trillion-parameter mixture-of-experts models, and while MoE activates only a slice of parameters per token, self-hosting a 1T flagship is out of reach for almost everyone without serious infrastructure, pushing most users toward hosted endpoints and undercutting the open-weights advantage. Benchmark placement is also often second-best among reasoning models, i.e. strong but chasing the closed frontier rather than leading it, and some Western adopters weigh provenance and data-transparency questions that come with any Chinese-lab model. The signal is credible and rising, but the models reward well-resourced teams far more than hobbyists.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Ant Ling (Bailing)'s official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Ant Ling (Bailing) unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Ant Ling (Bailing) directly →
Spotted something out of date? Suggest an update →
Alternatives to Ant Ling (Bailing)
More in LLMs & Chat
