OpenAI announces the rollout of its new GPT-5.3-Codex-Spark model, in collaboration with Cerebras, specializing in real-time software development. This highly responsive model, which operates at over 1,000 tokens per second, allows developers to receive instant feedback in live coding environments. Codex-Spark is engineered for agentic coding, enabling machines to autonomously handle tasks while keeping developers engaged in the iterative process. It surpasses its predecessor, GPT-5.1-Codex-mini, by offering faster task completion and refined capabilities in precise code editing, planning revisions, and contextual inquiries. The introduction of Cerebras’ Wafer-Scale Engine enhances the model’s performance in latency-sensitive projects. Codex-Spark is now available in a research preview for ChatGPT Pro users via the Codex app, CLI, and VS Code extension, with API access for select partners. The exciting developments from this partnership aim to bring ultra-fast inference capabilities to future AI workloads, targeting a broader AI landscape by 2026.
Source link
