A new OpenAI efficiency push could make GPT-6 much cheaper to use
imjustnewatai · x · 2026-07-28
The post argues that GPT-6 may end up much cheaper to use, even if it is not cheaper to train.
It points to reported OpenAI work that has already cut inference costs by more than 50% on some models, plus claims that GPT-5.5 improved serving speed by over 20%. The author’s thesis is that OpenAI is building the infrastructure runway for GPT-6: faster kernels, smarter scheduling, memory reuse, lower precision, caching, and model-assisted infra optimization.
The core prediction is not just cheaper tokens, but cheaper problem-solving: a larger model could still be affordable if it solves tasks in far fewer attempts than today’s agents.
More from Infra
- Kimi K3 reportedly keeps block attention residuals in a 93-layer design — burny_tech · 2026-07-28
- NASA’s new administrator says SpaceX-backed orbital data centers will happen — elonmusk · 2026-07-28
- Reddit users test whether unlocked CMP 170HX cards can power local AI rigs — Tritheone69 · 2026-07-28
- Open-weight models are pitched as a security win for large companies — jessi_cata · 2026-07-28
- Dolphin is being trained on Trinity Large Thinking 398B across 72 RTX 4090s — QuixiAI · 2026-07-28
- GitHub list catalogs public real-time datasets over HTTP and WebSockets — tom_doerr · 2026-07-28