AIBending Spoons reportedly uses self-hosted open-weight models for about 99% of AI requests and tokens
A post citing CEO Luca Ferrari says Bending Spoons uses frontier models for only about 1% of its AI work, reserving them for the most complex tasks or to check its open models.
TLDR
A post attributing the figures to Bending Spoons CEO and co-founder Luca Ferrari says the company self-hosts open-weight models for about 99% of its AI requests and tokens. Frontier models account for roughly 1%, handling the most complex work or checking the open models. The account credits this approach with keeping the company fully vendor-neutral and token costs extremely low.
Combined views
1 Source, first seen 2h ago
Bending Spoons reportedly uses self-hosted open-weight models for about 99% of AI requests and tokens
A post citing CEO Luca Ferrari says Bending Spoons uses frontier models for only about 1% of its AI work, reserving them for the most complex tasks or to check its open models.