OpenAI Says Its Jalapeño Chip Beats Nvidia

OpenAI says its own AI chip, Jalapeño, completes tasks more efficiently and returns answers faster than rival systems, according to benchmarks the company published in a blog post Tuesday.

In the tests, Jalapeño outperformed Nvidia's GB200 and GB300 superchips on AI inference — the work of running a trained model to answer a question or run an agent.

OpenAI's hardware vice president, Richard Ho, told reporters the chip offers the "best of both worlds" — lower latency and higher throughput — where AI systems typically "have to make a trade-off between the two."

The Numbers

OpenAI used InferenceX, a benchmarking platform for inference performance, and compared Jalapeño against the best results recorded at the time, which came from Nvidia's GB200 or GB300 superchips.

OpenAI says Jalapeño delivered 1.5 to 1.9 times more AI work per watt across the GPT-OSS 120B, DeepSeek R1 and Kimi K2.5 1T models.

It also reported 1.7 to 3.6 times lower end-to-end latency across those same three models.

That means "faster responses, more responsive agents, and more reliable access as the demand grows," Ho said.

Bloomberg reported the chip will let OpenAI customers choose models that offer either lower cost or faster answers.

Not a Breakup With Nvidia

Even with the benchmark wins, Ho said OpenAI does not expect to replace its entire chip lineup with Jalapeño, and that its overall compute strategy still includes "very good partners" like Nvidia.

OpenAI is already working on the second and third generations of the chip.

The rivalry runs both directions. The Daily Caller reported that Nvidia plans to spend billions of dollars to compete against China's AI models and against its own American AI customers.