Databricks

@databricks

Across seven held-out internal and external benchmarks: ↳ More than 2x faster than leading third-party and open-source models at comparable performance ↳ An average of 2.5 steps, and less than 6 seconds latency, even for the hardest tasks ↳ Pareto-optimal across a range of retrieval budgets
打开原帖#511482
  1. Frontier

    Nathan Lambert: wedding break — community curious but confused on AI
  2. Frontier

    Gary Marcus: live on OnPoint with MeghnaWBUR
  3. Frontier

    Ethan Mollick: GDP sims miss policy for explosive growth + displacement