OpenAI launched GPT-5.3-Codex-Spark on Cerebras chips for near-instant code generation at 1000+ tokens/sec. First major non-Nvidia inference partnership complements GPUs for low-latency. Available to Pro users via apps and extensions.
Key Points
- 1.Runs on wafer-scale Cerebras processors
- 2.>1000 tokens/sec speed
- 3.128K context, text-only
Impact Analysis
Diversifies OpenAI hardware amid Nvidia tensions. Enhances real-time coding flow for developers. Trades some capability for speed.
Technical Details
Optimized for low-latency inference. Underperforms full Codex on complex benches. Research preview for enterprises.
