Ant Group's Ling-3.0-flash-VL reaches the AI efficiency frontier with 5.5B active parameters
As seen on the 24/7 Wall St. homepage on September 11, 2026.
Ant Group is giving away weights that hit the efficiency frontier with just 5.5 billion active parameters, which is the kind of cost curve that squeezes pricing for closed US model vendors.
Ling-3.0-flash-VL, Ant Group’s new flash tier open weights model, scores 25 on the Artificial Analysis Intelligence Index. At 124B total parameters 5.5B active parameters, it sits on the Intelligence vs. Active Parameter Pareto Frontier @AntLingAGI has released https://t.co/XjfdrMhf0H
- Replies1
- Reposts0
- Likes25
Continue ReadingShow less
Ant Group released Ling-3.0-flash-VL as open weights, putting a model that activates only 5.5 billion parameters at inference directly in the hands of builders who would otherwise pay a closed US vendor. Artificial Analysis places it on the Intelligence vs. Active Parameter Pareto Frontier, so no publicly tracked model at that compute cost scores as well. Enterprises now have a self-hostable option at the flash tier.
Ignoring this means missing a direct pricing threat to closed vendors in the cost-sensitive, high-volume segment.
Sponsored
Are You Ready To Retire, Or Years Behind?
Most Americans suspect they're behind on retirement and never find out. Advisor.com's free matching tool pairs you in about three minutes with a vetted fiduciary advisor who can help you with investing, taxes, retirement, estate planning, and more. No minimums. No sales call. Find out where you stand.
That index score of 25 is the single benchmark Artificial Analysis used to position the model. Sitting on the Pareto Frontier is a meaningful designation: it signals that users get the most intelligence per unit of compute at this active-parameter tier, rather than paying for extra parameters that do not translate into measurably better outputs.
For closed model vendors, an open-weights competitor that matches or approaches their efficiency puts direct pressure on pricing. Enterprises evaluating inference costs now have a self-hostable option at the flash tier, which is typically the cost-sensitive, high-volume segment of the market. That is the competitive dynamic worth tracking as adoption data begins to emerge.