GPT-5.2-Codex

Pricing — 1 offering(s)

Showing the active price and any recorded history. Full pricing history is available via the paid API — see API docs.

Capability profile

Operator guidance

Superseded by gpt-5.3-codex at the same input/output pricing (though gpt-5.2-codex is not formally deprecated — see limitations). modelglass-coding's own scores show gpt-5.3-codex materially ahead on SWE-bench Verified (78.00% vs. this model's 72.40%, both via Vals AI's independently-run leaderboard) — prefer gpt-5.3-codex for any new agentic-coding integration. This entry exists for completeness and for callers still pinned to this snapshot.

Use cases

• Agentic coding tasks in Codex or similar environments — OpenAI's own stated purpose for this model
• Existing deployments pinned to gpt-5.2-codex specifically — superseded by gpt-5.3-codex, though not formally deprecated

Limitations

• Superseded in capability by the newer gpt-5.3-codex (OpenAI's current flagship agentic coding model) — but gpt-5.2-codex is NOT deprecated: it does not appear on OpenAI's deprecations page and its model docs page carries no lifecycle warning (re-checked 2026-09-03; an earlier 'deprecated per OpenAI's docs' note here was incorrect)
• Only the v1/responses endpoint is supported — Chat Completions, Realtime, Assistants, and Batch are not available for this model, per OpenAI's own docs
• Max input capped at 272,000 tokens — lower than the 400,000-token context-window figure alone would suggest
• Capability ratings above are qualitative vendor framing, not a single cited benchmark run — see modelglass-coding for this model's actual sourced SWE-bench Verified/Pro scores (72.40%/41.04% respectively)

Citations

• GPT-5.2-Codex model spec (OpenAI's own docs)
• OpenAI API pricing
• OpenAI models