Ant's Ling-3.0-flash-VL Scores 25 on AA Index With Just 5.5B Active Params
ArtificialAnlys · x · 2026-09-12
Ant Group released Ling-3.0-flash-VL, an open-weights (MIT) reasoning model adding image/video understanding to Ling-3.0-flash. Artificial Analysis findings:
- MoE: 124B total, 5.5B active per token, 256K context, text/image/video input
- Scores 25 on the AA Intelligence Index, on the Intelligence vs. Active Params Pareto frontier; Qwen3.5 122B A10B scores 16, Mistral Medium 3.5 15
- Low hallucination: 14% AA-Omniscience accuracy, 22% hallucination rate (Inkling Small: 33% accuracy but 63% hallucination); factual recall is weak
- Weak on hard agentic tasks: 16% AutomationBench-AA, 0% Terminal-Bench v4.0
- Verbose: 50k output tokens per task vs 30k for Inkling Small, with cost implications
Related event: Ant's Ling-3.0-flash-VL Hits Pareto Frontier with Just 5.5B Active Params(2 posts)→
More from Models
- ToneBench eval: GPT-6.1 Terra ranks #23 at writing — cheaper, not better, than GPT-5.6 Terra — OnlyProggingForFun · 2026-10-28
- Cursor user complains it silently switches to Grok, which fails basic tasks — kylegawley · 2026-10-28
- Musk says Grok Bot keeps improving, calls Vault feature a game changer — beffjezos · 2026-10-28
- Claude refuses to gather legitimately public information, sparking guardrail backlash — mkheck · 2026-10-28
- DesignArena launches Code to Video Arena: models direct 15-45s films purely in code — yshan2u · 2026-10-28
- ChatGPT's 'yes, but with an important clarification' often precedes pure nonsense — Crixusgannicus · 2026-10-28