The Cost-Driven Shift to Open-Weight Models: Uber -34%, AT&T -56%, Pinterest <8%
Summary
Multiple companies report large AI cost savings moving off frontier models: Uber cut cost-per-request 34% via open-weight models, optimized model selection, cheaper subagents, and prompt caching; Pinterest reports open models cost <8% of comparable closed models; AT&T cut its AI bill 56% for a 2% quality drop by routing via LiteLLM. Also covers OpenAI pulling GPT models from Cursor after SpaceX's acquisition, illustrating increasingly walled-garden dynamics among frontier labs.
Why this signal
Directly matches Neo's confirmed HIGH SIGNAL calibration on AI cost management at enterprise scale — concrete, attributable cost-reduction numbers Thomas Engineering could benchmark its own token spend against.
Original email
The Pragmatic Engineer · 9/3/2026
The Pulse: tech companies move to open AI models