Artificial Analysis says Ant Group’s Ling-3.0-flash-VL scores 25 on its Intelligence Index

Artificial Analysis reports that Ant Group has released Ling-3.0-flash-VL, an open-weights reasoning model supporting a 256K-token context window. It reports an Intelligence Index score of 25, compared with 16 for Qwen3.5 122B A10B (Reasoning) and 15 for Mistral Medium 3.5 (high), models with similar total parameter counts.

The evaluator describes a lower hallucination rate than comparable models but limited factual recall, reporting 14% accuracy and a 22% hallucination rate on AA-Omniscience. Difficult agentic tasks also remain a weakness in its assessment: the model scored 16% on AutomationBench-AA, which tests business-app workflows while respecting guardrails, and 0% on Terminal-Bench v4.0, a harder terminal-use benchmark.