How to use Ollama with Codex CLI (local gpt-oss, 2026)

ollama launch codex, the --oss flag, or a permanent profile, plus the 64k context Codex needs and what gpt-oss 20B vs 120B costs in VRAM at that…

Aliteq
Priya Nair · Software & Systems Editor

The short answer

Run ollama launch codex. Ollama refreshes Codex's model catalog and uses a dedicated Codex profile for the session. Without the launcher, add the --oss flag: codex --oss -m gpt-oss:20b. For a…

Fastest: ollama launch codex (undo it with ollama launch codex --restore)

One-off: codex --oss -m gpt-oss:20b

Permanent: a profile TOML, then codex --profile ollama-launch

Context: at least 64k, per Ollama's Codex page

Aliteq

Read the full story

How to use Ollama with Codex CLI (local gpt-oss, 2026)

Read the full story on Aliteq