NeurIPS 2026 paper 'Intelligence per Watt' measures on-device AI efficiency across 1M queries
HazyResearch · x · 2026-10-20
The paper Intelligence per Watt: Measuring Intelligence Efficiency of Local AI (Hazy Research/Stanford, presented with SnorkelAI) is accepted to NeurIPS 2026; authors Jon Saad-Falcon and Avanika Narayan present it at a Nov 4 reading group.
Across 20+ models, 8 accelerators, and 1M real-world queries, IPW measures how much capability local AI delivers per watt — and which queries belong on-device instead of in the cloud.
More from Infra
- Citi projects $5.35 trillion in cumulative capex from Amazon, Microsoft, Google and Meta through 2030 — Beth_Kindig · 2026-10-20
- DisaggKD paper splits teacher and student across nodes to lift KD memory limits — Ruixuan Tu · 2026-10-20
- Free 44-session GPU kernel engineering series: naive GEMM to FlashAttention on H100 and Blackwell — Winter_Mistake_3185 · 2026-10-20
- Meta details Prometheus gigawatt supercluster: 85%+ pretraining goodput for Muse — AIatMeta · 2026-10-20
- Where Should Enterprise AI Agents Run? Cloud, On-Prem, Edge or Hybrid — DavidLinthicum · 2026-10-20
- REAX inference engine runs Perplexity Decider up to 6x faster, free for individuals — airesearch12 · 2026-10-20