Ant Ling Previews Ling-3.1-flash 1019 Update with Open Weights Coming Soon
Ant Ling announced an upcoming 1019 checkpoint update for the Ling-3.1-flash model, bringing benchmark improvements and a pending open weights release. API access and open weights are confirmed as coming soon without a specific release date. This update signals ongoing performance gains for a prominent large language model, which will soon be accessible for local deployment via open weights. It benefits developers and researchers looking for improved agentic capabilities and cost-effective execution. The 1019 version raises the model to 44 on the Artificial Analysis Intelligence Index and improves agent evaluations, increasing AutomationBench-AA from 62% to 68% and AA-Briefcase to 1524 Elo. The model features 560 billion total parameters with approximately 25 billion activated per token.
## BACKGROUND
Ling-3.1-flash is a hybrid reasoning mixture-of-experts language model developed by InclusionAI, an initiative under Ant Group. Artificial Analysis evaluates AI models across various performance metrics, while benchmarks like AutomationBench-AA measure a model's competency in executing complex agentic SaaS workflow tasks.