GPT-5 token cost & pricing

GPT-5 is OpenAI's flagship reasoning model and the default choice when a task involves multi-step logic, code generation or anything where a wrong answer is expensive. It carries a 400K context window and supports adjustable reasoning effort, which means the same model can answer a quick classification question or spend a minute working through a hard math problem.

GPT-5 specifications

What GPT-5 costs in practice

Four common workloads, priced at list rates. Monthly figures assume 30 days.

How many requests does $100 buy?

Assuming a 1,000-token prompt and a 500-token response per request.

How GPT-5 pricing actually works

At $1.25 per million input tokens and $10 per million output, GPT-5 sits in the middle of the frontier tier — cheaper than Claude Opus 4.1 on both sides, more expensive than Gemini 2.5 Pro. The 8x input-to-output ratio is the number that matters: if you generate long responses, output dominates your bill. Cached input at $0.125 per million is 90% off list, so any application reusing a large system prompt should enable caching before anything else.

When to use GPT-5

Reach for GPT-5 when correctness matters more than latency: agentic coding, data analysis, complex extraction, multi-document synthesis. Avoid it for simple classification, sentiment tagging or short chat replies — GPT-5 mini handles those at roughly a fifth of the input price and a fifth of the output price, and the quality difference is usually invisible on easy inputs.

How to cut your GPT-5 bill

Two levers move GPT-5 costs the most. First, trim the system prompt — every redundant few hundred tokens is paid on every request. Second, route: send low-confidence or simple cases to GPT-5 mini and escalate only when needed. Teams that route typically cut spend 50-70% versus running everything on the flagship.

Other OpenAI models

Cheaper alternatives to GPT-5

• Mistral Large 2Mistral29% cheaper
• Qwen3 MaxAlibaba36% cheaper
• Claude Haiku 4.5Anthropic47% cheaper
• Kimi K2Moonshot72% cheaper