Ant Ling
32 Intel
Ant Ling AGI is the account for Ant Group's AGI initiative, publishing the Ling and Ring families of open-source large language models built on a mixture-of-experts architecture. The Ling series are general-purpose models and the Ring series are reasoning-focused, with releases such as Ring-1T reaching trillion-parameter scale and performing strongly on IMO-style reasoning benchmarks. The group also publishes the multimodal Ming model family for image, text, audio, and video.
Is this you? Sign in with X to claim this profile.
Intel
Ant Ling open-sources Ling-3.1-flash-1019 plus a Ling-3.1-flash-base checkpoint. It reports an Artificial Analysis Intelligence Index of 44 (up from 41), with gains on AA-Briefcase, AutomationBench-AA, CritPt and SciCode. Weights are on Hugging Face and ModelScope.
Ant Ling describes Ring-2.5-1T-Zero, a 1T model trained from base with RL and no extra human annotation. It reports format rewards becoming redundant at scale, a four-stage pipeline, and 100K traces distilled into Qwen-32B beating 800K from DeepSeek-R1. Thread only, paper not reviewed.
Ant Ling confirmed Ling-3.0-flash-fin is actually a 124B-parameter MoE model with 5.1B active parameters despite the "flash" name, and noted an fp4 quantization is available for local deployment.
Ant's inclusionAI group made Ling-2.6-1T, a trillion-parameter instruct model, available on OpenRouter. It uses a fast-thinking approach it says cuts token cost by about 75% while holding up on AIME26 and SWE-bench Verified. Claims are vendor-reported and unverified.
Ant Ling open-sources Ling-3.1-flash, a roughly 560B MoE with 25B active parameters and 1M context. The thread covers MXFP4/FP8 quantization for 4-GPU serving, Artificial Analysis speed and price figures, and front-end generation rankings. Benchmarks are vendor-cited.
Ant Group details FinFIRST, a finance-agent benchmark built with 50+ finance professionals using atomic rubrics to grade both answers and evidence quality, benchmarking 15 model configurations under a unified search/browse/Python setup.
Ant Ling open-sources Ling-3.0-flash-Sante, a healthcare-tuned MoE (124B total, 5.1B active) with reported gains in deep research, multi-agent collaboration and tool use. Benchmarks are self-reported and the model card was not readable.
Ant Ling open-sources Ring-2.6-1T, a trillion-scale thinking model with high and xhigh reasoning levels, trained with asynchronous RL via IcePop. Self-reported results include 74.00 SWE-bench Verified, 95.83 AIME 26 and 66.18 ARC-AGI-V2 pass@2, plus coding-agent demos.