Verified, then logged.
Sun 22 Nov 2026 · 00:46 KST / 15:46 UTC한국어 | ENSearch /

Breaking01 / AI

OpenAI's Pando backend cuts median push latency by about 80%

An engineer, seen from behind, faces a monitor showing a blurred code branching graph, with long rows of storage server racks visible through a glass wall. AI-generated image

OpenAI said on 2 Nov 2026 that Pando, its Git-compatible backend, cut median push latency by about 80%. The figure is company-published, and Pando supports the whole of OpenAI's single large code repository.

Throughput. A push is the operation that transfers code. According to OpenAI's technical write-up, Pando serves about 1 petabyte of Git data a day and handled more than 57 terabytes in one hour in one instance. Both are company-published figures.

Release scope. OpenAI said it is contributing improvements to the Git client to open source, through public repositories and the upstream Git project. These contributions are separate from a general release of Pando itself.

Git, the system Pando is compatible with, is a distributed version control system. The documentation in the openai/git repository on GitHub states that Git is an open-source project covered by the GNU General Public License version 2 (GPLv2).

OpenAI has not disclosed whether Pando itself will be made generally available, or on what schedule.