Full Breakdown
OpenAI Launches GPT-5.3-Codex-Spark on Cerebras Chips
2/13/2026, 5:50:59 AM
Introduction of GPT-5.3-Codex-Spark
On February 12, 2026, OpenAI unveiled its latest coding model, GPT-5.3-Codex-Spark, which operates on Cerebras hardware, marking a significant shift in the company's approach to AI model deployment. This new model is designed for rapid code generation, achieving speeds of over 1,000 tokens per second—approximately 15 times faster than its predecessor, GPT-5.1-Codex-mini. In comparison, Anthropic's Claude Opus 4.6 reaches about 68.2 tokens per second in its standard mode, highlighting the competitive landscape in AI coding tools.
Technical Specifications and Performance
GPT-5.3-Codex-Spark is tailored for speed rather than depth, focusing on coding tasks rather than general-purpose applications. It features a 128,000-token context window and is currently available to ChatGPT Pro subscribers through various platforms, including the Codex app and VS Code extension. OpenAI's benchmarks indicate that Spark outperforms older models on software engineering tasks, although independent validation of these claims has not been provided.
Partnership with Cerebras
The integration of Cerebras' Wafer Scale Engine 3, a chip boasting 4 trillion transistors, is central to the performance of Codex-Spark. OpenAI's partnership with Cerebras, announced in January 2026, is part of a multi-year agreement valued at over $10 billion. This collaboration aims to enhance AI responsiveness and enable real-time collaboration for users, positioning Spark as a productivity tool for rapid prototyping.
Official Statements & Responses
OpenAI emphasized that Codex-Spark represents a significant milestone in its relationship with Cerebras, stating, "Integrating Cerebras into our mix of compute solutions is all about making our AI respond much faster." Sean Lie, CTO of Cerebras, expressed enthusiasm for the collaboration, noting, "What excites us most about GPT-5.3-Codex-Spark is partnering with OpenAI and the developer community to discover what fast inference makes possible."
Criticism & Opposition
Despite the promising specifications, some industry observers have raised concerns regarding the lack of independent validation for OpenAI's performance claims. Previous tests indicated that OpenAI's models, when run on Nvidia hardware, delivered significantly lower speeds, with GPT-4o achieving only 147 tokens per second. This discrepancy raises questions about the reliability of the benchmarks presented by OpenAI.
What's Next
As OpenAI continues to roll out GPT-5.3-Codex-Spark, the company is expected to explore new interaction patterns and use cases that leverage the model's fast inference capabilities. The ongoing development and potential IPO plans signal a commitment to expanding its influence in the AI landscape.
Verbatim Quotes
- “Cerebras has been a great engineering partner, and we’re excited about adding fast inference as a new platform capability,” — Sachin Katti, Head of Compute at OpenAI
- “This preview is just the beginning.” — Sean Lie, CTO and Co-founder of Cerebras
