Unofficial calculator

Percent in. Human estimate out.

GPT-5.6 SolUpdated July 2026Max / Ultra / Extra High

Codex Usage Calculator

Estimate Codex tasks from the shared OpenAI agentic pool. GPT-5.6 Sol, Terra and Luna are visible beside the ChatGPT and Work experiences.

Build your estimate

Pick a platform, match the window it shows, then read the likely range.

Model: GPT-5.6 SolClaude Fable 5 supportedMax / Ultra / Extra High

Usage setup

Match what your provider shows before reading the estimate.

Live estimate

Choose your setup

Start with the product and model, then add effort or agent mode when it matters.

Remaining allowance

Time until reset

Your estimate

Plus5 hours

Around 65 Codex tasks left

Based on 65% remaining in the 5 hours window.

Used 35%Remaining 65%

Remaining

65%

Used

35%

Likely range

29-101 Codex tasks

Estimate

65 Codex tasks

Safe pace

6-20 Codex tasks/hour

Typical API reference per task

$0.03-$0.21 (GPT-5.6 Sol)

Status: NormalReliability: Medium-low

You still have reasonable margin, but avoid wasting usage.

Main factors

Codex uses the shared OpenAI agentic pool and token-based model rates, adjusted for reasoning, execution mode, repo size and task complexity.

Limit basis: ChatGPT chat uses the closest normalized message window. Codex, Work and workspace agents share an agentic usage and credit pool, with token-based model rates and optional Max or Ultra execution modes.

Codex, ChatGPT Work and workspace agents draw from the same agentic usage and credit pool when available on your plan.

Agentic usage is token-based. Task counts are normalized equivalents, not a fixed message cap.

GPT-5.6 Sol is rolling out gradually on eligible plans.

This tool is not affiliated with OpenAI, Anthropic, Google, Perplexity, Cursor, Windsurf or Devin. Results are unofficial estimates based on the remaining percentage and options you enter. Real limits can vary by plan, model, feature, system capacity, context length, files, task complexity and provider changes. API cost references are not subscription charges.

Popular setups

Jump straight to the estimate you need

These are the setups people most often compare when a provider shows a percentage instead of a simple count.

GPT-5.6 / CodexSol reasoning, Extra High, Max and Ultra agent modes.Claude Max / Fable 5Pro, Max 5x, Max 20x and task-aware Fable 5 estimates.Gemini UltraAI Pro, Ultra tiers and Deep Think.Cursor UltraAgent, background and Max Mode estimates.

From Percent To Prompts

The calculator turns a vague remaining percentage into an easier estimate of prompts, messages, coding tasks or agent runs left.

Start With Percent Left

Use the remaining percentage your AI tool shows, not the percentage already used.

Match The Window

Choose the reset window, plan and platform so the estimate uses the closest preset.

Read Human Units

See estimated Codex tasks, ChatGPT messages, Claude messages or similar usage labels.

FAQ

Quick answers about how the estimator treats remaining percentage, reset windows and unofficial AI usage ranges.

Why can Codex estimates vary?

Codex usage can vary a lot by reasoning effort, repo size, local or cloud execution and task complexity.

Does this calculate exact tokens?

No. It estimates prompts, tasks, messages or usage units from your remaining percentage.

Why is it approximate?

AI platforms use dynamic limits based on plan, model, system capacity, context, files and task complexity.

What does remaining percentage mean?

It is the percentage the platform shows as remaining, not the percentage already used.

Is this official?

No. It is an independent estimator.

Other calculators

This tool is not affiliated with OpenAI, Anthropic, Google, Perplexity, Cursor, Windsurf or Devin. Results are unofficial estimates based on the remaining percentage and options you enter. Real limits can vary by plan, model, feature, system capacity, context length, files, task complexity and provider changes. API cost references are not subscription charges.