Using Claude Code extensively - solid tool, but need the $100 tier for real dev work.

Testing OpenAI's Codex as comparison. Anyone know the context length for GPT-5 through Codex? (practical limit, not necessarily the full 400K)