OpenAI posts Jalapeño benchmarks before production ramps up
OpenAI published Jalapeño's first results on August 25. Across three public models, it reported 1.5 to 1.9 times more AI work per watt. It also reported 1.7 to 3.6 times lower end-to-end latency. OpenAI plans to deploy the chip internally by year end.
These are company benchmarks. SemiAnalysis watched runs in OpenAI's lab and described the chips as engineering samples. It said production is scheduled to ramp up in 2027 and warned that the comparisons need caveats.
Read article →