OpenAI shows Doom running on custom inference chip ported via Codex

BenBajarin · x · 2026-08-26

OpenAI's custom inference chip, "Jalapeño," is designed as a generalized chip capable of running various models and workloads. During lab tests with OpenAI engineers, the chip ran the InferenceX benchmark. As a demonstration of its capabilities, OpenAI showed the game Doom running on the chip, which was ported entirely using Codex prompts. This versatility increases the likelihood of Microsoft adopting the chip internally through their IP sharing agreement.

Original post →

More from Infra

Infra channel →