Source
The full stack behind abundant intelligence
OpenAI is implementing an integrated compute strategy across hardware and software, recently sharing performance results for its first custom inference chip, Jalapeño.
Why it matters
Increased efficiency in chips and software can lower costs for users, making complex AI tasks more economically practical for businesses to implement.
The details
- Jalapeño outperformed commercial systems in throughput per kilowatt and token latency.
- OpenAI partners with providers including NVIDIA, AWS, AMD, and SoftBank for compute.
- GPT-5.6 Sol used 54% fewer output tokens on the Coding Agent Index.
Get the weekly recap
The stories like this one, picked and explained — once a week, straight to your inbox.