Ant releases Ling-3.2-flash: 560B total/25B active multimodal model beating Opus 5 on vision benchmarks
Aiden_Tech_Ai · x · 2026-11-07
Ant Ling has introduced Ling-3.2-flash, a natively multimodal model with 560B total / 25B active parameters.
Key capabilities: stronger complex visual understanding, long-horizon multimodal agents, 3D understanding & interaction, and medical intelligence.
In release evaluations it outscores Opus 5 (max) on several benchmarks:
- OmniDocBench1.5: 92.21 vs 90.56
- ZEROBench(Sub): 43.71 vs 39.52
- Vision2Web: 85.86 vs 80.20
More from Models
- OpenAI appears to sit on models longer: internal checkpoints lead releases by months — haider1 · 2026-11-07
- Xbow tests GPT-6 Cyber vs GPT-6.1 Sol on 65 vulnerable apps: faster, not broader — moyix · 2026-11-07
- NLP researcher observes LLMs are changing the meaning of 'similarly' — yoavgo · 2026-11-07
- Power user burns 700M-1.8B tokens daily on Codex, says $200 plan no longer enough — kimmonismus · 2026-11-07
- Hesamation: using ChatGPT feels like traveling back to 2024 AI — Hesamation · 2026-11-07
- OpenAI merges usage across products, Chat-only Plus users fear hitting 5-hour limits — br_web · 2026-11-07