Smart glasses test: Octen search at 60ms, model generation is the bottleneck; Mercury 3 wins
Scobleizer · x · 2026-10-20
KZouAPT shared latency tests from building a Sense API for smart glasses: Octen search took a median of 60ms, with model generation as the real bottleneck. Among fast models tested, Mercury 3 delivered the best speed-quality balance, with 0.9s median generation and 2.2s end-to-end response. Context: Stefano Ermon announced Mercury 3 as the most intelligent diffusion LLM released to date, outperforming GPT-6 Luna and Haiku 5.5 on quality while running at 1000+ tok/sec.
More from Embodied
- Week 2: migrating a 2D Python game to Unreal 3D with Claude and Tripo, 680 tests included — Dead_Fox_7 · 2026-10-20
- Maker Arm + Dream-1: give robots zero-shot tasks from your browser in plain language — gharik · 2026-10-20
- Dream-1 claimed to be first model to complete the difficult "Stand Up Bottles" robotics task — gharik · 2026-10-20
- Viral rant: will humanoid robots gate walking upstairs behind a $19.99 subscription? — VraserX · 2026-10-20
- Driver crashes Tesla after forgetting he'd disengaged FSD while posting on X — PTrubey · 2026-10-20
- MIT embeds symmetry into neural network to fix fetal brain MRI, E(3)-Pose presented at ECCV 2026 — MIT News AI · 2026-10-20