AI/TLDR

GPT-5 Pro

Parallel test-time compute for the hardest answers.

Overview

GPT-5 Pro is OpenAI's highest-capability tier of the GPT-5 family, launched on 2025-08-07 alongside GPT-5 itself. It is a variant of GPT-5 that, in OpenAI's words, 'thinks for ever longer' by using scaled but efficient parallel test-time compute to produce the most comprehensive and accurate answers. It first shipped inside ChatGPT for Pro subscribers, then became available in the API as the gpt-5-pro model on 2025-10-06.

Because GPT-5 Pro spends far more compute per query than standard GPT-5, it targets research-grade and high-stakes problems where a better answer is worth the extra latency and cost. On launch it set a new state of the art on GPQA Diamond, scoring 88.4% without tools — the best result in the GPT-5 family at the time. It accepts text and image input and returns text.

GPT-5 Pro has since been superseded by later Pro tiers (GPT-5.1 Pro, GPT-5.2 Pro, and beyond), so this original August 2025 release is now a legacy model. Its specs and benchmarks are documented here for reference.

Released2025-08-07
LicenseProprietary
WeightsAPI only
ParametersUndisclosed
Context400K
Max output128K
ArchitectureUndisclosed (reasoning model with parallel test-time compute)
ModalitiesText, Vision
StatusDeprecated

Benchmarks

  1. GPQA Diamond (no tools)88.4%
  2. AIME 2025 (with Python tools)100%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

Pricing

Input$15.00 / 1M tokens
Output$120.00 / 1M tokens

API price for the gpt-5-pro model (available from 2025-10-06).

Pricing source ↗

Strengths

  • Highest answer quality in the original GPT-5 family via parallel test-time compute
  • State-of-the-art GPQA Diamond (PhD-level science) score of 88.4% without tools at launch
  • Strong competition-math performance (100% on AIME 2025 with Python tools)
  • Large 400K total context window with 128K max output
  • Accepts image input in addition to text

Best for

  • Research-grade science, math, and finance problems where answer quality outweighs cost
  • High-stakes professional work (legal, medical, technical analysis) needing the most thorough reasoning
  • Difficult coding and debugging tasks that benefit from extended deliberation
  • Verifying or stress-testing answers from faster, cheaper models

How to access

ProviderModel ID
OpenAI API ↗gpt-5-pro

GPT Pro — every version

The full lineage of the GPT Pro line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.

VersionReleasedContextLicense
GPT-5.5 Procurrent2026-04-231.05MProprietary
GPT-5.4 Pro2026-03-05Proprietary
GPT-5.2 Pro2025-12-11Proprietary
GPT-5.1 Pro2025-11-19Proprietary
GPT-5 Pro2025-08-07Proprietary

FAQ

What is GPT-5 Pro?

GPT-5 Pro is the highest-capability tier of OpenAI's GPT-5 family, released on 2025-08-07. It is a variant of GPT-5 that uses scaled, efficient parallel test-time compute to think longer and deliver the most comprehensive and accurate answers on hard problems.

How is GPT-5 Pro different from GPT-5?

GPT-5 Pro runs the same underlying GPT-5 model but spends much more test-time compute per query, trading higher latency and cost for better answer quality. At launch it scored 88.4% on GPQA Diamond without tools, the best in the GPT-5 family.

How much does GPT-5 Pro cost in the API?

In the OpenAI API, the gpt-5-pro model is priced at $15.00 per 1M input tokens and $120.00 per 1M output tokens — roughly 12x the output price of standard GPT-5. It became available in the API on 2025-10-06.

Is GPT-5 Pro still the latest model?

No. GPT-5 Pro (August 2025) is a legacy model that has been superseded by later Pro tiers such as GPT-5.1 Pro, GPT-5.2 Pro, and newer releases. Check OpenAI's current model list for the most up-to-date Pro tier.