Back to today’s briefing

Story perspectives

OpenAI Unveils GPT-5.3-Codex-Spark for Non-Nvidia Hardware

2/13/2026

45 4 Full Breakdown

Summary

Story summary

OpenAI released the GPT-5.3-Codex-Spark coding model on February 12, 2026, the first production AI compatible with non-Nvidia hardware.
Cerebras' Wafer Scale Engine 3 powers the model and enables over 1,000 tokens per second.
Codex-Spark targets rapid coding tasks and is available to ChatGPT Pro subscribers.
OpenAI credits its Cerebras collaboration for faster response times despite independent benchmarks showing Nvidia hardware slower.

1 of 1

Story summary

OpenAI released the GPT-5.3-Codex-Spark coding model on February 12, 2026, the first production AI compatible with non-Nvidia hardware.
Cerebras' Wafer Scale Engine 3 powers the model and enables over 1,000 tokens per second.
Codex-Spark targets rapid coding tasks and is available to ChatGPT Pro subscribers.
OpenAI credits its Cerebras collaboration for faster response times despite independent benchmarks showing Nvidia hardware slower.