Codex logo

How much does Codex cost to run?

The honest cost breakdown for Codex: how you actually pay, live model prices, and realistic monthly scenarios.

Codex has the friendliest cost story of the vendor agents: it runs on ChatGPT subscription plans, so if you already pay for ChatGPT, Codex's marginal cost is effectively zero until you hit plan limits. The alternative is the OpenAI API, where you pay per token at the model prices below.

On the API side the model spread is wide: GPT-5.4 Mini at $0.75 per million input tokens, GPT-5.4 at $2.50, and GPT-5.5 at $5. Because agent work is input-heavy, that choice changes a heavy month by multiples.

The table below prices Codex's models against three usage tiers, computed from live prices in our database.

How you actually pay
  • ChatGPT subscription: the default. Codex use is bundled into the plans most people already have, with limits that scale by tier. The best per-dollar deal if you subscribe anyway.
  • OpenAI API: per-token billing at the prices below, for automation, CI, and usage-based team setups.
  • Model routing: GPT-5.4 Mini for routine steps, GPT-5.4 for everyday work, GPT-5.5 for the hardest tasks. Codex supports switching, and the price spread makes routing worth the setup.
What a month costs, per model

Estimated monthly model cost for three usage tiers, computed from live per-token prices. Agent work is input-heavy (the model re-reads context every step), so the tiers assume roughly 10 input tokens per output token: a few tasks a day: 5m input + 0.5m output tokens daily; steady all-day use: 25m input + 2.5m output tokens daily; agent runs continuously: 100m input + 10m output tokens daily.

  • OpenAIGPT-5.5
    Input$5/M
    Light$1,200/mo
    Daily driver$6,000/mo
    Heavy / always-on$24,000/mo
  • OpenAIGPT-5.4
    Input$2.5/M
    Light$600/mo
    Daily driver$3,000/mo
    Heavy / always-on$12,000/mo
  • OpenAIGPT-5.4 Mini
    Input$0.75/M
    Light$180/mo
    Daily driver$900/mo
    Heavy / always-on$3,600/mo

Estimates cover model tokens only (no subscriptions, caching discounts, or batch pricing) and assume list prices, which the daily sync keeps current. Your real usage will differ; treat the tiers as brackets, not predictions.

Frequently asked questions
Is Codex free with a ChatGPT subscription?

Codex is included with ChatGPT plans: usage is bundled until you reach your tier's limits, so for existing subscribers the marginal cost of agent work is effectively zero. Heavy users step up a tier or supplement with API billing when they hit caps.

How much does Codex cost on the API?

Per-token at OpenAI's model prices: GPT-5.4 Mini at $0.75 per million input tokens, GPT-5.4 at $2.50, and GPT-5.5 at $5, with output costing more per token. An agent-heavy day burns tens of millions of input tokens, so the scenario table on this page is the realistic way to estimate a month.

Codex or Claude Code: which costs less?

Whichever vendor you already subscribe to: both bundle agent use into their plans, and bundled beats per-token for interactive work. On pure API pricing OpenAI is somewhat cheaper at the workhorse tier (GPT-5.4 at $2.50 versus Sonnet 4.6 at $3 per million input tokens). See our Claude Code vs Codex comparison for the capability side.

What is the cheapest way to run a Codex-style agent?

If subscription limits frustrate you, the open-source route is the cost floor: an agent like OpenCode or Cline with a budget model (DeepSeek V4, GLM 4.7 Flash) through OpenRouter runs all day for under a dollar. You trade OpenAI's polish for model freedom and lower bills.

Picking a model first? See the best models for Codex with benchmarks and prices side by side.

Share:
Details:

  • Agent

    Codex

  • Software cost

    Subscription or API

  • Models priced

    3

  • Updated

    June 2026