OpenAI Says Jalapeño Beats Nvidia GB300 On 2 Key AI Metrics

OpenAI says Jalapeño beat Nvidia GB300 in two inference benchmarks at 700 watts, with deployment planned later in 2026 to cut costs and speed responses.

Key Points:

OpenAI says Jalapeño led Nvidia’s GB300 in throughput per unit of power and response speed during testing.

The 700-watt inference chip is designed to lower data center costs while serving AI models faster.

Jalapeño is not built for model training and was not tested against Nvidia’s newer Vera Rubin generation.

OpenAI Benchmarks

OpenAI said its custom Jalapeño processor outperformed Nvidia’s GB300 in tests measuring AI work per unit of power and the speed of returning responses.

Richard Ho, OpenAI’s chip chief, said Jalapeño combines high throughput with low latency, potentially letting customers choose models that prioritize lower costs or faster answers.

Developed with Broadcom, Jalapeño achieved its test results while operating at 700 watts, according to Ho. Power is a major data center cost.

Also Read: RWA Market Surges 48.7%, But One Token Explains Almost Everything

Nvidia Context

The comparison has clear limits.

Jalapeño was not tested against Nvidia’s newer Vera Rubin generation, which has started shipping, and the OpenAI chip is designed for inference rather than model training. Nvidia remains a key supplier.

OpenAI also uses Cerebras Systems technology for some inference workloads, particularly with smaller models, while Ho said Jalapeño can handle larger ones. “We have so much need for compute,” he said.

Public testing used one smaller OpenAI open-source model alongside models from DeepSeek and Moonshot AI, with broader advantages appearing on Moonshot’s larger Kimi model. A second version is already far along, with tape-out expected in the coming months, while OpenAI has started conceptual work on a third generation.

The chip follows OpenAI’s Jun. 24 announcement with Broadcom, when the companies said Jalapeño moved from design to manufacturing tape-out in nine months with help from OpenAI models. That was the first generation.

Alexey Bondarev

Alexey Bondarev is the Head of Content at Yellow.com, having reported on crypto for the last 10 years. He specializes in in-depth Research and Learn pieces, with a focus on analytical reporting, industry context, and the bigger forces shaping crypto, from the AI era and security technologies to fintech innovation. He believes that everything digital will imminently overcome everything analogue and is working hard to make that come true.