GPT-5.2

GPT-5.2 is OpenAI's previous frontier chat model. The current model docs still sell the gpt-5.2 snapshot, but they now recommend GPT-5.6 Sol for new work. This page used to treat Instant / Thinking / Pro as the live catalog.

Key Features

• Previous frontier, still on the API: Snapshots gpt-5.2 and gpt-5.2-2025-12-11.
• 400,000-token context, 128,000 max output, knowledge cutoff 2025-08-31.
• Reasoning effort: none (default), low, medium, high, xhigh.
• Modalities: text + image in, text out.
• Successor: docs point new builds at GPT-5.6 Sol.

Use Cases

• Pinned production traffic that cannot move off gpt-5.2 yet.
• Long briefs that fit in 400k but do not need GPT-5.6 Sol's 1.05M window.
• Vision-in, text-out workflows that already work on this snapshot.

Limitation: launch-era Instant ($0.75 / $5) and Pro ($10 / $80) prices are stale unless you re-verify them on a live pricing page. Do not plan new agents around Instant / Thinking / Pro as if they were the current API list.

Pricing

From the 2026-08-17 model page:

Getting Started

1. Read the GPT-5.2 model page.
2. Call gpt-5.2 or pin gpt-5.2-2025-12-11.
3. Set reasoning.effort only if you need more than the default none.
4. For new systems, compare GPT-5.6 Sol before you lock the old snapshot.

Frequently Asked Questions

Is GPT-5.2 still the flagship?

No. It is the previous frontier. Docs recommend GPT-5.6 Sol.

Are Instant, Thinking, and Pro still the API catalog?

Not on the current model page. That page sells one GPT-5.2 snapshot and a reasoning-effort knob.

Should I keep the launch blog as the canonical URL?

This entry now points at the live developers docs, which carry current limits and price.

Alternatives

• GPT-5.6 Sol: Current OpenAI frontier.
• Claude Opus 4.5: Anthropic frontier if you are leaving the GPT line.
• Kimi K3: 1M-context open-weight alternative.

Tips

1. Treat $1.75 / $14 as the verified snapshot price, not Instant or Pro.
2. Default effort is none. Turn it up only when the task needs it.
3. Confirm the model page before you quote SWE-Bench or GDPval numbers from the launch post.

Conclusion

GPT-5.2 is still a 400k-context, $1.75 / $14 API model, but it is no longer the model OpenAI tells you to start with. Read the official model docs, then decide whether to stay pinned or move to GPT-5.6 Sol.