OpenAI Unveils Jalapeño Custom Inference Chip Benchmarks at Hot Chips 2026
OpenAI's first Jalapeño benchmarks, shown at Hot Chips 2026, claim 1.5x to 1.9x more AI work per watt and 1.7x to 3.6x lower latency than Nvidia GB200 and GB300 systems on three open models. Here is how it was tested, what SemiAnalysis checked, and what is still missing.





