AI Lab @AntGroup, we envision AGI as humanity's shared milestone. Our Language Model @AntLingAGI and LLaDA, Embodied AI @robbyant_brain, OSS projects @AReaL_AI.
Introducing LingBot Touch ๐ฅ A JEPA based tactile foundation model for robotic manipulation is now open sourced!
๐ขโ A shared, scalable architecture supporting cross-sensor, cross-hand, cross-task, and cross-action-model transfer
๐โ Real-robot validation across two dexterous hands, 26 manipulation tasks, and four action models
๐โ Tiny and Small variants for lightweight, high-frequency tactile inference
๐ฅโ Nearly 5,000 hours of tactile pretraining data, with performance improving as the data scales.
What changes when robots can feelโnot just see? ๐คโ
Meet LingBot-Touch, our open-source tactile foundation model towards dexterous manipulation, pretrained on nearly 5,000 hours of real tactile data.
Through unified heterogeneous tactile modeling, LingBot-Touch supports transfer across sensors, dexterous hands, tasks, and action models.
Watch it in action. ๐ #LingBot #EmbodiedAI #Robotics
Seeing details is only the beginning. Connecting them is the next step.
๐ Introducing Zooming without Zooming v2 (ZwZ2).
Our EvidenceConnect framework generates questions that require connecting visual evidence across regions within a single image.
Trained with RLVR on just 20.4K of these examples, ZwZ2-4B/9B/27B improve over their corresponding base models on all 25 benchmarks:
โข ZwZ2-4B: 49.41% โ 54.79%
โข ZwZ2-9B: 51.63% โ 56.59%
โข ZwZ2-27B: 55.52% โ 60.51%
The gains extend to multi-image long-context understanding, embodied spatial reasoning, and multimodal agent tasksโeven with single-image training.
๐ https://t.co/yhBke8bFM3
๐ ๏ธ https://t.co/54U8LXdttw
๐คhttps://t.co/YfaBvyuYiS
โก Introducing UBTree: a frontier dLLM-based method for inference speedup๐
On Ling3-Flash-124B, UBTree delivers ~1,010 output tokens/s with 4-GPU tensor parallelism (TP4) at concurrency 1, averaging a 3.43ร end-to-end speedup over target-only autoregressive decoding across six benchmarks.
UBTree uses a unigram proposer to provide candidate draft tokens and a bigram selector to connect these candidates into a tree. We build this draft tree completely in parallel.
Today we are releasing two drafters for the @AntLingAGI family models:
๐ Ling-3.0-tiny-8B: https://t.co/LIybyZsUT2
๐ Ling-3.0-flash-124B: https://t.co/KE0p2imaot
Use them to accelerate your Ling-3.0 now!
More draft models are coming over the next month.
๐ Paper: https://t.co/5sQ0NAS5SO
๐ Website: https://t.co/xUeqRSoUbJ
โก๏ธโก๏ธโก๏ธLing-3.1-flash-1019 is now open source! Further post-training for reasoning and real-world agent workflows.
Along with another release, Ling-3.1-flash-base, an open foundation for further training.
Download, evaluate and build with @AntLingAGI.
๐ Ling-3.1-flash-1019
HF: https://t.co/GVEBPei2gD
ModelScope: https://t.co/Zgm20ZKkZV
๐ Ling-3.1-flash-base
HF: https://t.co/eIwg60AEQ9
ModelScope: https://t.co/g7HEgERRSN
Introducing @AntLingAGI Ling-3.1-flash-1019 ๐
It scores 44 on the Artificial Analysis Intelligence Index, up from 41 for Ling-3.1-flash. Gains in agentic work: AA-Briefcase ~1400โ1524 Elo and AutomationBench-AA 62%โ68%.
API access and open weights coming soon. Stay tuned๐
๐ค Meet our healthcare-enhanced MoE model, @AntLingAGI Ling-3.0-flash-Sante! Built for real-world healthcare research and workflows, with strengthened capabilities in deep research, multi-agent collaboration, and tool use.
Sante shows a significant improvement on professional healthcare tasks, and provides stronger complex medical reasoning.
๐ Open source and open weights:
๐ https://t.co/XW1bhljyH7
๐ https://t.co/rOj2QxQD5K
Introducing @AntLingAGI Ling-3.1-flash-1019 ๐
It scores 44 on the Artificial Analysis Intelligence Index, up from 41 for Ling-3.1-flash. Gains in agentic work: AA-Briefcase ~1400โ1524 Elo and AutomationBench-AA 62%โ68%.
API access and open weights coming soon. Stay tuned๐
Thanks to everyone at InclusionAI Tech Night during #OpenSourceAIWeek. We talked medical benchmarks, usable open models, and agent systems in the real world.
Thanks to our wonderful speakers and community partners for making this night happen! @sgl_project@open_intel_fdn@ZenMuxAI
SGLang-Omni supports @TheInclusionAI's Ming-3.0 on day 0!
๐๏ธ Ming-3.0-asr-tiny โ Chinese/English ASR with dialect support
๐ Ming-3.0-tts-tiny โ voice cloning in 7 languages
๐ง Ming-3.0-audio-flash โ 126B-A6.5B MoE for audio understanding and reasoning
Cookbooks๐
Meet the Ming-3.0 speech families:
โ Ming-3.0-tts for stories, narration & voice assistants
โ Ming-3.0-audio for meetings, interviews & voice-driven services, offering a more efficient deployment option
โ Ming-3.0-asr for customer calls, voice search & domain terms. 4.41% average error across four ASR categories.
Try out on Hugging Face & ModelScope! https://t.co/sUTJPdes4b
#opensource #inclusionAI
Build for real work, Ling-3.1-flash is now open source. ~560B total params, ~25B active/token, up to 1M-token context.
Across work, โจ๏ธcoding & ๐ชhealthcare: 1,673 Elo on GDPVal-AA v2.1, 75.16 on FrontierSWE, and 65.35 on HealthBench Professional.
๐คHugging Face: https://t.co/ukHHd3dxo1
๐คModelScope: https://t.co/klkZXExtHE
#opensource #inclusionAI #LLM
Ling-3.1-flash is now open source: ~560B total parameters, ~25B active per token and up to 1M context.
Built for real work, it can turn a rough brief into a page, mobile UI or interactive prototype teams can review and refine. See the demo.
The FIRST Robbyant Embodied AI Challenge is live. ๐ค๐
We're inviting builders worldwide to push embodied intelligence forward โ real hardware, real environments, real impact. ๐ฆพ
Be part of history in the making. Prizes. Recognition. The frontier. ๐ Learn more & register โ https://t.co/6IfjmLYyWG
#EmbodiedAI #Robotics
Speaker lineupโs here ๐ More names are on the way! Join InclusionAI Tech Night during #OpenSourceAIWeek for candid talks on open models, AI benchmarks, and real-world agents, plus Token Shots and networking!
๐ San Francisco
๐ Oct 18 ยท 4โ8 PM
๐ https://t.co/9g6pzl6kam
Thanks to the SGLang community for joining InclusionAI Tech Night as a Community Partner! ๐
See you in San Francisco on Sun, Oct 18 for candid conversations on open models, AI benchmarks, and agent systems.
RSVP: https://t.co/9g6pzl6kam
Where to find #SGLang at Open Source AI Week:
Sun Oct 18 ยท InclusionAI Tech Night, SF
Tue Oct 20 ยท SGLang x LinkedIn Meetup, Sunnyvale
Tue Oct 20 ยท AI Infra Night at PyTorchCon, San Jose
Links below. https://t.co/5WNhtAaQ6E @open_intel_fdn#OpenSourceAIWeek#OpenIntelligence
1๏ธโฃ Sun Oct 18 ยท InclusionAI Tech Night: Your AI Reality Check in SF, hosted by Ant Open Source and inclusionAI. Request to join: https://t.co/W3Gr71TzhD
2๏ธโฃ Tue Oct 20 ยท SGLang x LinkedIn Meetup in Sunnyvale. Talks on LLM serving, including how LinkedIn runs SGLang in production. Limited spots, request to join: https://t.co/yaVAZ3BwfN
3๏ธโฃ Tue Oct 20 ยท AI Infra Night #PyTorchCon in San Jose with dstack and Crusoe. GPU and infra folks, after the conference day. Request to join: https://t.co/6pe0pGzpGK
Can we trust what the AI benchmarks says? Can an agent system be smarter than its best model?
Join inclusionAI at #OpenSourceAIWeek 2026 for tech talks, happy-hour drinks, food, and Token Shots ๐ธ
Bay Area ยท RSVP: https://t.co/9g6pzl6kam