← All models

GPT-5.3-Codex-Spark

Published Publisher OpenAI

Overview

GPT-5.3-Codex-Spark is a smaller version of GPT-5.3-Codex designed for low-latency, real-time coding in Codex. OpenAI released it as a research preview on February 12, 2026. (openai.com)

Capabilities and uses

  • Supports interactive coding workflows such as targeted edits, restructuring logic, refining interfaces, planning projects, and translating files. Users can interrupt or redirect it while it works.
  • OpenAI reports generation speeds above 1,000 tokens per second on Cerebras hardware and strong performance on SWE-Bench Pro and Terminal-Bench 2.0 relative to its task-completion time.
  • Its default working style favors minimal, targeted edits; it does not automatically run tests unless asked.
  • It runs on Cerebras Wafer Scale Engine 3 hardware. (openai.com)

Availability and limitations

At launch, the model was text-only with a 128k-token context window. OpenAI announced a research-preview rollout to ChatGPT Pro users through the Codex app, CLI, and VS Code extension, alongside API access for a small group of design partners. Preview usage had a separate rate limit, and high demand could lead to restricted access or temporary queues. These are launch conditions and should not be assumed to describe current availability. (openai.com)

Pricing

As checked on November 16, 2026, a model-specific API price in USD for input, cached input, and output tokens could not be verified in OpenAI's published API pricing. OpenAI's Enterprise token-based rate card lists the model as a research preview rather than publishing numerical input or output rates. Do not substitute the price of GPT-5.3-Codex, which is a different model. (developers.openai.com)

Official sources