OpenAI logo

OpenAI · Language Model

GPT-5.2

OpenAI's flagship for professional knowledge work.

4.6Released December 2025Instant · Thinking · Pro tiers

Overview

GPT-5.2 is OpenAI's most capable model series to date, released in December 2025 in three variants — Instant, Thinking and Pro. The launch followed Sam Altman's internal "code red" memo and was widely read as OpenAI swinging back at the base model layer after Google's Gemini 3 had landed several punches. Where late-2025 ChatGPT updates focused on product surface area (browsing, shopping, UI polish), GPT-5.2 targets the underlying intelligence: long-horizon reasoning, tool reliability, long context and professional "artifact" work like spreadsheets, slides and structured documents.

The three tiers map to different jobs: Instant handles fast everyday tasks and writing, Thinking adds deliberate step-by-step reasoning, and Pro is the heaviest variant for the hardest problems. A sibling release, GPT-5.2-Codex, is further optimized for agentic coding with stronger long-horizon refactors and migrations. Reviewers describe GPT-5.2 as a genuine step up in vision and agentic tool-calling, even if independent benchmarks paint a more nuanced picture than OpenAI's own charts.

Key capabilities

  • Three tiers (Instant, Thinking, Pro) to trade speed for depth
  • Stronger agentic tool-calling and long-context understanding
  • Improved vision and document/artifact generation
  • GPT-5.2-Codex sibling tuned for large code changes

At a glance

Intelligence rank

Frontier tier (top 10)

Input price

~$1.75 / 1M tokens

Output price

~$14 / 1M tokens

Pricing: Premium frontier pricing (~$1.75 in / $14 out per 1M tokens).

Pros & cons

What we like

  • Best-in-class reasoning for professional knowledge work
  • Reliable tool use and long-context handling
  • Codex variant excels at real-world software engineering

Trade-offs

  • Among the most expensive models on the market
  • Benchmark leads are narrower than the marketing suggests
  • Pro tier latency is high for everyday tasks

The verdict

4.6Aggregate review

GPT-5.2 is OpenAI's "code red" counterpunch and it mostly lands. It is the safe default for high-stakes reasoning, agentic workflows and polished document output — provided you can absorb frontier pricing. For routine chat, the Instant tier or a cheaper rival usually does the job.

Best for: Complex reasoning, agentic tool use and professional document work.

Frequently asked questions

What's the difference between GPT-5.2 Instant, Thinking and Pro?

Instant is fastest and best for everyday writing and Q&A; Thinking adds deliberate step-by-step reasoning for harder problems; Pro is the heaviest, most thorough variant for the toughest tasks at the cost of speed.

Is GPT-5.2 good for coding?

Yes. The base model is strong, and GPT-5.2-Codex is purpose-built for agentic coding — large refactors, migrations and long-horizon software tasks.

How much does GPT-5.2 cost?

It sits in the frontier price band, roughly $1.75 per million input tokens and $14 per million output tokens — among the priciest models available.

Further reading

Related models