OpenAI introduced GPT-5.3-Codex-Spark for interactive software development, emphasizing generation speed and a 128,000-token context window. At launch, access was offered as a research preview to ChatGPT Pro users.

Context

Speed changes the rhythm of small edits: request a change, inspect it, and refine the result. But useful speed is the time to a working solution, not merely the time until text appears. If a faster model needs more corrections or misses a constraint, the generation advantage can disappear during review. Both parts belong in a comparison.

Sources & authors

  1. Introducing GPT-5.3-Codex-Spark
    OpenAI · February 12, 2026
OpenAImodels