OpenAI has announced the first benchmark results for Jalapeño, its first custom-designed inference chip, reporting significant gains in AI performance, power efficiency and response times across several large language models.
CEO of OpenAI Sam Altman announced on X, "we made a chip and it is fast."
we made a chip and it is fast
— Sam Altman (@sama) August 25, 2026
ALSO READ: Canada Slaps Up To 50% Tariffs On $20-Billion US Imports As Trade War Worsens
According to OpenAI, Jalapeño delivered 1.5 to 1.9 times more AI work per watt and 1.7 to 3.6 times lower end-to-end latency than the comparison systems across GPT-OSS 120B, DeepSeek R1 670B and Kimi K2.5 1T. For highly interactive workloads, performance was between 2.1 and 4.1 times higher.
The chip is rated for 700 watts, although OpenAI said sustained measured power remained at or below 550 watts during testing. On Kimi K2.5 1T, Jalapeño achieved about 1.5 times higher peak performance per watt and 3.4 times lower end-to-end latency than the comparison system.
OpenAI said Jalapeño was designed specifically for modern AI inference, with its hardware, memory, networking and software developed together to reduce data movement and communication delays. The architecture also keeps model state, including the KV cache, closer to the compute resources handling inference.
Artificial intelligence itself played a role in developing the chip. OpenAI said AI-assisted design helped the team reach tapeout in nine months, while also optimising parts of the chip's arithmetic circuits.
The company further used Codex with GPT-Astra to optimise three open-weight models within two months. Selected AI-generated implementations for GPT-OSS attention and mixture-of-experts blocks were 1.5 to 1.8 times faster than existing human-written versions.
ALSO READ: Zoom Gives Lukewarm Outlook Despite Expanded Product Line
OpenAI plans to begin deploying Jalapeño in its compute infrastructure by the end of 2026, with Gen 2 already under development and Gen 3 taking shape. The company said it will continue using accelerators from NVIDIA and other partners for both training and inference alongside its own chips.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
