OpenAI Code Interpreter

license

Proprietary

service

proprietary(OpenAI-managed sandbox)

source

closed

isolation

undisclosed

access

API/tool only(no self-host, no open runtime)

Code Interpreter is a proprietary managed service. OpenAI releases no runtime and offers no self-hosted version, and it documents no isolation mechanism beyond calling each container a sandboxed virtual machine.

• https://developers.openai.com/api/docs/guides/tools-code-interpreter recorded 2026-09-29

  Code Interpreter is reached only through the OpenAI API as a container: "A container is a fully sandboxed virtual machine that the model can run Python code in." The body carries no license text for the tool; its only self-hosted and open-source strings are site navigation labels ("Self-hosted sandboxes" for the Agents API, "Codex for Open Source"), and no Code Interpreter runtime or source is offered.

Code Interpreter ships inside ChatGPT's advanced data analysis feature and the Responses, Chat Completions and Assistants APIs, riding ChatGPT's mass-market weekly audience. OpenAI discloses no standalone usage figure for the tool itself, so adoption is read conservatively from that surface, on the assumption that only a fraction of ChatGPT users invoke code execution.

• https://developers.openai.com/api/docs/guides/tools-code-interpreter recorded 2026-09-29

  The guide, subtitled "Allow models to write and run Python to solve problems.", publishes no usage figure for the tool; searched the body for weekly users, monthly users, active users and customers: none.

• https://en.wikipedia.org/wiki/ChatGPT recorded 2026-09-29

  The ChatGPT article states that "ChatGPT reached 900 million weekly active users" in February 2026, the surface Code Interpreter ships in.

Code Interpreter is a Python-only sandbox that lets a model retry failing code, with containers of up to 64 GB of memory, but each container is ephemeral and OpenAI documents neither its isolation technology nor its scaling internals. That narrower, less transparent design keeps it behind general-purpose sandboxes that isolate a wider range of runtimes.

• https://developers.openai.com/api/docs/guides/tools-code-interpreter recorded 2026-09-29

  The guide says the tool lets models "write and run Python code in a sandboxed environment" and that "A container expires if it is not used for 20 minutes", with memory tiers of "1g" (default), "4g", "16g" or "64g".

• https://developers.openai.com/api/docs/pricing recorded 2026-09-29

  The pricing page lists "1 GB $0.03, 4 GB $0.12, 16 GB $0.48, 64 GB $1.92 per 20-minute session per container." for Hosted Shell and Code Interpreter.