Wispr Flow's in-house Canto 1 speech model tops voicearena dictation benchmark
Scobleizer · x · 2026-11-16
Scobleizer argues we spent 40 years learning to type and the next 10 are about computers learning to listen. Wispr Flow hired Ariya Rastrow — who grew Amazon's Alexa speech team from 0 to 400 — as CSO four months ago, and shipped its own speech model, Canto 1, two months ago.
Results since launch: desktop users fix 13% fewer dictations, slow dictations down 54%, daily dictated words up 7%; this week voicearenaai ranked Wispr Flow #1 on its dictation benchmark. His thesis: voice wins not by understanding you but by earning enough trust that you stop fixing its output — and companies building their own models will own how we talk to machines.
More from Models
- Claude's formulaic creative writing sparks Reddit debate on editing AI output — GIMBruhBoobies · 2026-11-16
- User claims GPT-6 'Astra' can already control robots and fine-tuned an open model's weights — xmercury_one · 2026-11-16
- Swarm orchestration runs 12 parallel agents to build a Blender 3D model, claims tester — venturetwins · 2026-11-16
- Gemini Robotics 2.5 debuts with 96% pick-and-pack success on Mercedes-Benz lines — dl_weekly · 2026-11-16
- Google solicits feedback on 'Gemini 4 Argon, Carbon, Hydrogen' in apparent leak — deliprao · 2026-11-16
- James Douma: Grok is currently the best tool for understanding what's happening in the world — jamesdouma · 2026-11-16