Introducing GPT-5.3-Codex-Spark

2026-02-02 · Source: OpenAI News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

OpenAI has released GPT-5.3-Codex-Spark, a smaller, ultra-fast version of GPT-5.3-Codex, designed for real-time coding. This model, optimized for low-latency hardware, delivers over 1000 tokens per second and is the first milestone in OpenAI's partnership with Cerebras. It features a 128k context window and is text-only, available as a research preview to ChatGPT Pro users and select API design partners. Codex-Spark excels in interactive work, making targeted edits without automatically running tests, and demonstrates strong performance on SWE-Bench Pro and Terminal-Bench 2.0 benchmarks. The deployment also includes end-to-end latency improvements across the request-response pipeline, benefiting all models by reducing overhead and time-to-first-token.

Key takeaway

For Machine Learning Engineers and developers focused on real-time coding workflows, you should explore GPT-5.3-Codex-Spark to significantly reduce iteration times. Its ultra-low latency, powered by Cerebras hardware, enables more fluid interaction and rapid prototyping, potentially transforming how you approach targeted code edits and interface refinements. Consider integrating it into your development environment for tasks where immediate feedback is paramount.

Key insights

GPT-5.3-Codex-Spark offers ultra-fast, real-time coding capabilities via specialized hardware and optimized inference.

Principles

• Latency is critical for interactive AI collaboration.
• Specialized hardware can complement general-purpose GPUs.

Method

The model uses a persistent WebSocket connection and optimized inference stack to reduce client/server roundtrip overhead by 80%, per-token overhead by 30%, and time-to-first-token by 50%.

In practice

• Utilize Codex-Spark for rapid iteration in coding tasks.
• Expect separate rate limits for specialized hardware models.

Topics

• GPT-5.3-Codex-Spark
• Real-time Coding
• Cerebras WSE 3
• Low-latency Inference
• Software Engineering Benchmarks

Best for: Machine Learning Engineer, CTO, VP of Engineering/Data, Software Engineer, AI Engineer, AI Chatbot Developer

Related on AIssential

• OpenAI launches GPT-5.3-Codex-Spark for ultra-fast real-time coding — GPT-5.3-Codex-Spark leverages Cerebras' WSE3 chip for ultra-low latency AI inference in real-time coding.
• Cerebras says its chips run a trillion-parameter AI model nearly 7 times faster than GPU clouds — Cerebras' wafer-scale architecture delivers unprecedented speed for trillion-parameter AI inference, outperforming GPU clouds significantly.
• Introducing GPT‑5.3‑Codex‑Spark — GPT-5.3-Codex-Spark prioritizes speed for real-time coding, enabling more productive iterative workflows.
• OpenAI Releases a Research Preview of GPT‑5.3-Codex-Spark: A 15x Faster AI Coding Model Delivering Over 1000 Tokens Per Second on Cerebras Hardware — Specialized hardware and optimized connections dramatically accelerate AI coding model inference.
• OpenAI has yet another new coding model and this time it's really fast — OpenAI's new Codex-Spark model prioritizes real-time coding speed over raw accuracy using specialized Cerebras hardware.
• OpenAI GPT-5.3-Codex-Spark Shows What's Possible With Cerebras — OpenAI's new coding model, GPT-5.3-Codex-Spark, leverages Cerebras hardware, challenging Nvidia's dominance and expanding AI hardware options.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by OpenAI News.