OpenAI retires speed-king Codex-Spark after just 7 months

OpenAI's Codex lead announced GPT-5.3-Codex-Spark, its fastest model at over 1,200 tokens per second, will be retired next week after just 7 months, citing declining usage and stronger successors. Commenters noted its high-throughput legacy with anticipation.

2026-09-12 ~ 2026-09-12 · 2 related posts

• OpenAI to retire GPT-5.3-Codex-Spark next week, speculation swirls on Astra successor — imjustnewatai · 2026-09-12
• OpenAI Retires Codex-Spark, Its Fastest Model, After Just 7 Months — 新智元 · 2026-09-12