Amazon saves millions by replacing Sonnet 4.6 with Strands Decider

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by psnehanshu on Tuesday November 10, 2026 @02:50AM (#50437307)

Ok, but why did they chose Sonnet in the first place? Couldn't they have chosen Haiku or GPT Luna or maybe even some small cheap Chinese model?

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by theshrike79 on Tuesday November 10, 2026 @03:07AM (#50437421)

[delayed]

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by psnehanshu on Tuesday November 10, 2026 @05:26AM (#50438331)

I think for this purpose a small model will do reasonably well

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by jeremyjh on Tuesday November 10, 2026 @07:43AM (#50439550)

Right. So we can either assume the engineers at Amazon are fucking clowns who don't do any engineering, or we can assume they tested and found that Sonnet was the cheapest model that met their reliability targets.

If you think the former is a likely hypothesis, then why did you even click on TFA?

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by psnehanshu on Tuesday November 10, 2026 @07:49AM (#50439614)

Let's keep the discussion civil

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by Hamuko on Tuesday November 10, 2026 @03:22AM (#50437519)

Not sure if it would even need to be a small Chinese model. Amazon would surely have the necessary hardware and infrastructure to run a bigger Chinese model with a low per-token cost too.

[ Parent | on HN ]

Re:Amazon saves millions by replacing Sonnet 4.6 with Strand... by loafoe on Tuesday November 10, 2026 @03:58AM (#50437746)

Strands decider actually uses a small cheap/free Chinese model for its base architecture: Qwen

[ Parent | on HN ]