Alex Smola's Santa Clara company says Higgs Realtime Duplex processes voice and video in 160 ms chunks while handling tools and delegating harder questions.
By [RuntimeWire Staff](https://runtimewire.com/author/runtimewire-staff)
· Published
Primary source: [Aligned News — AI Intelligence](https://x.com/rdesh26/status/2118684296441442347)
Why it matters #
Full-duplex interaction could let a voice agent listen while speaking, watch a screen, call tools and hand hard problems to a bigger backend model without dead air. Boson AI's target is for an hour with its 45B front model to cost at most one-tenth as much as an hour of a person's time.
Boson AI, co-founded and led by Alex Smola, introduced Higgs Realtime Duplex on November 5th, 2026, describing it as the next generation of Higgs Realtime and its first full-duplex interactive omni model. The model listens, watches and talks at the same time while rendering a face of its own, according to Boson AI. It also thinks, calls tools and delegates work to a larger model without pausing the conversation, Boson AI says.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.