Codex Pricing 2026: Free, $8 Go, $20 Plus, $100 Pro, Business Plans, Credits, Usage Limits

TL;DR

Codex is included in every ChatGPT plan: Free $0, Go $8/mo, Plus $20/mo, Pro from $100/mo with 5x or 20x limits, Business $20/user/mo on annual billing ($25 monthly), and Enterprise or Edu at custom pricing. There is no standalone Codex subscription. Usage is metered in credits. One credit costs 4 cents. On Plus, the published allowance is 10-100 GPT-5.6 Sol local messages per 5-hour window, 25-200 on Terra, and 250-2,000 on Luna.

An API key skips the windows and bills per token. GPT-5.6 Luna is $0.20/M input and $1.20/M output after the July 30, 2026 price cut. Terra is $2.00/$12.00. Sol is $5.00/$30.00. A multi-file refactor costs $2.40 on Sol and 10 cents on Luna.

Codex Plans and Prices

Codex ships inside ChatGPT Free, Go, Plus, Pro, Business, and Enterprise. Six tiers, plus an API key that needs no plan at all:

Business, Edu, and Enterprise workspaces on flexible pricing can purchase workspace credits. Plus and Pro users buy credits individually. Free and Go cannot buy credits at all, which makes Plus the real floor for anyone who hits a limit mid-task.

Codex Plus: What $20 Per Month Buys

There is no product called Codex Plus. The tier is ChatGPT Plus at $20/month, and Codex is included in it. That $20 covers Codex on web, CLI, IDE, and iOS at one shared allowance.

• Local messages: 10-100 per 5-hour window on GPT-5.6 Sol, 25-200 on Terra, 250-2,000 on Luna.
• Cloud chats and code reviews: separate published allowances against the same five-hour window.
• Credit top-ups: available. Plus and Pro users who reach a limit can buy credits instead of upgrading the plan.
• Weekly limits: apply on top. OpenAI states that additional weekly limits may apply and publishes no numbers for them.

The ranges are wide because the limits are credit-based underneath. A one-line fix sits near the bottom. An agentic run that reads thirty files and writes ten sits near the top. Cheaper models get more messages because they burn fewer credits per token.

Codex Pro: The $100 and $200 Tiers

Pro no longer means $200. It starts at $100/month, and you pick a multiplier at checkout. The multiplier applies to the Plus allowance exactly, with no rounding:

• Pro 5x, $100/mo: 50-500 GPT-5.6 Sol local messages per window, 125-1,000 on Terra, 1,250-10,000 on Luna.
• Pro 20x, $200/mo: 200-2,000 on Sol, 500-4,000 on Terra, 5,000-40,000 on Luna.

$100 buys 5 Plus-units of allowance. $200 buys 20. Twice the price for four times the capacity. If you hit Pro 5x limits even occasionally, 20x is the cheaper marginal capacity.

Codex Usage Limits

OpenAI publishes per-model allowances in three columns: local messages per 5 hours, cloud chats per 5 hours, and code reviews per 5 hours. Local messages and cloud chats share one five-hour window. The published local-message allowances on the GPT-5.6 lineup:

Legacy models are still listed with their own allowances on Plus: 15-80 local messages on GPT-5.5, 20-100 on GPT-5.4, and 60-350 on GPT-5.4 mini. GPT-5.4 and GPT-5.4 mini leave Codex for ChatGPT sign-in users on August 31, 2026. Both stay available through the API and in Codex sessions authenticated with an API key.

The weekly limit is real and unpublished

OpenAI's pricing tables carry one footnote under every plan: additional weekly limits may apply. No numbers accompany it. That is the honest state of the documentation, and it is why community estimates of the weekly cap vary so widely. Treat the five-hour allowances as the only figures OpenAI stands behind.

The 5-hour window was temporarily removed in July

On July 12, 2026, Thomas Sottiaux, who leads Codex at OpenAI, announced on X that the company was temporarily removing the 5-hour usage limit restriction for all Plus, Business, and Pro plans. No end date was published, and OpenAI's pricing documentation still lists the five-hour tables as policy. The weekly limits stayed in place throughout.

The removal followed sustained complaints that the five-hour framing did not survive contact with a real repository. On the OpenAI developer forum, one Plus user documented the gap:

The mechanism behind the complaint is arithmetic. Limits are enforced in credits, and context-loading at the start of a session burns credits before the first edit. A 300k-token repo read costs 37.5 credits on Sol, or $1.50, against an allowance that a light user stretches across a whole afternoon. Plan for the window to return.

Codex Credits: Burn Rates Per Model

Credits are the unit Codex meters usage in. OpenAI publishes the exact burn rate per million tokens for every model:

One credit costs 4 cents. Cross-reference any row against the API price list and the conversion falls out. Sol costs $5.00 per million input tokens at the API and 125 credits in Codex. Terra costs $2.00 and 50 credits. Luna costs $0.20 and 5 credits. Every row holds at exactly 4 cents. Credits are API token pricing in different units.

Three consequences follow from the table:

• Luna burns a twenty-fifth of Sol on input and output alike. That is why Luna gets 250-2,000 Plus messages per window against Sol's 10-100.
• Output burns 6x the input rate on every model. Test generation and scaffolding drain a window far faster than debugging and code review.
• Cached input burns a tenth of fresh input. Long sessions in one repo benefit. Jumping between repos resets the advantage. Cache writes on 5.6-generation models bill at 1.25x the uncached input rate.

What a Codex Session Costs

The credit arithmetic is exact, so per-task costs are derivable from token volume. Input tokens times 125 plus output tokens times 750, per million, at 4 cents per credit, gives the Sol figure:

Token volumes are estimates. The arithmetic is not. Running the same work on Luna costs a twenty-fifth as much, which is the single largest lever on a Codex bill and the one most people never pull. Test generation costs disproportionately more than its input size suggests because output burns 750 credits per million on Sol against 125 for input.

This also explains why two developers on the same plan report completely different experiences. One running small fixes gets 80 messages per window. One running 300k-token agentic refactors gets 10.

Codex API Pricing

Running Codex with an API key, or calling the models directly, bills per token with no subscription. Rates after the July 30, 2026 cut:

GPT-5.6 reached general availability on July 9, 2026 at $5/$30 for Sol, $2.50/$15 for Terra, and $1/$6 for Luna. Three weeks later, OpenAI cut Terra 20% and Luna 80% while holding Sol flat. Luna input is now 25x cheaper than Sol input. Cached-input reads keep the 90% discount on every model, and cache writes on 5.6-generation models bill at 1.25x the uncached input rate.

Codex CLI Pricing

Codex CLI is free software. You pay through one of two channels:

• ChatGPT sign-in: included on every plan, Free and Go included. CLI usage draws from the same local-message allowance as Codex on web and IDE, so a heavy CLI session and a heavy web session compete for one budget.
• API key: per-token billing at the rates above. No window limits, no weekly cap, no cloud features. Cost scales linearly with usage.

The CLI also supports custom model providers through config.toml (base_url, env_key, wire_api), so it can point at any Responses-compatible endpoint. Setup is in our Codex provider configuration guide.

Business and Enterprise

Business is $20/user/month on annual billing, or $25/user/month monthly. Its per-seat Codex allowance matches Plus exactly: 10-100 Sol local messages per window, 25-200 on Terra, 250-2,000 on Luna.

What Business and Enterprise add over individual plans is cloud features, GitHub, Slack, and Linear integrations, and admin controls. Business, Edu, and Enterprise workspaces on flexible pricing can also purchase workspace credits rather than per-seat top-ups. Earlier in 2026 Business offered pay-as-you-go Codex seats that billed for actual token usage. As of June 24, 2026, new pay-as-you-go seats are no longer available. Existing seats are unaffected.

Is Codex Free?

Yes. Codex is included in ChatGPT Free at $0/month, and Codex CLI works on Free with ChatGPT sign-in. OpenAI describes the Free tier as enough to explore Codex on quick coding tasks, which is an accurate description of its ceiling.

Two caveats. Free carries the lowest allowance of any plan. Free also cannot buy credits, so hitting a limit means waiting for the reset rather than paying to continue. Go at $8/month is the cheapest paid step up and is described as suitable for lightweight coding tasks. Plus at $20/month is the first tier that can buy credits mid-task.

There is no separate student price for Codex. Students use the same Free, Go, or Plus plans. Edu pricing covers institutions, not individuals.

Codex vs Claude Code Pricing

The subscription ladders mirror each other almost exactly. API pricing is where they diverged in July:

At matching sticker prices the subscription decision comes down to capability per dollar. On the API side the July 30 cut moved Terra below Anthropic's Sonnet tier on both input and output, and Luna at $0.20/$1.20 is the cheapest option on either side by a wide margin. Sol remains the most expensive flagship of the two on output. Full breakdowns are in Claude Code pricing, Codex vs Claude Code, and Claude vs ChatGPT.

Codex Subscription vs API Key

A Codex subscription buys a flat monthly bill and a hard ceiling. An API key buys no ceiling and a variable bill. The crossover depends entirely on which model you run.

• On Sol: sessions cost $0.40 to $2.40, so Plus ($20/mo) breaks even at roughly 10 to 50 sessions a month and Pro 5x ($100/mo) at roughly 40 to 250.
• On Luna: sessions cost 2 to 10 cents. Plus does not break even until several hundred sessions a month.

The API route fits developers who batch work into occasional deep sessions, teams building products on the models, and anyone who wants gpt-5.6-luna at $0.20/$1.20 for high-volume simple work. What you give up is cloud features, GitHub, Slack, and Linear integrations, and the predictable flat bill. What you gain is no five-hour window and no weekly cap.

Which Codex Plan to Pick

Free or Go ($0-$8/mo)

You want to evaluate Codex CLI or use it a few times a week. Free works for evaluation. Go at $8/month is the cheapest plan with usable headroom. Neither can buy credits when a limit lands.

Plus ($20/mo)

You code regularly but not in all-day agentic sprints. 10-100 Sol local messages per window covers most individual workflows, and 250 to 2,000 on Luna covers almost any of them. Credit top-ups handle the occasional overage without a plan change.

Pro 5x ($100/mo)

You hit Plus limits weekly. 50-500 Sol local messages per window, or 1,250-10,000 on Luna. This is the tier for daily multi-file agentic work.

Pro 20x ($200/mo)

You run parallel agents or sustained refactors. 200-2,000 Sol messages per window. Twice the price of Pro 5x for four times the capacity, so heavy users get better unit economics here.

Business or Enterprise

Your team needs integrations and admin controls. Business is $20/user/month annual, $25 monthly, at the Plus allowance per seat. Enterprise adds custom contracts.

API key

You run under 10 to 50 Sol sessions a month, you want Luna at $0.20/$1.20, or you need programmatic access with no window limits. For uncapped open-weight coding models, DeepSeek V4 Flash on Morph is $0.139/$0.278 per million tokens.

Pricing History

Skip the Usage Cap: Pay-Per-Token Open Models

If the recurring pain is the cap and not the price, the fix is a per-token API with no five-hour window and no weekly cap. Morph serves open-weight coding models on custom kernels through an OpenAI-compatible endpoint at https://api.morphllm.com/v1. You pay for the tokens you use, and nothing throttles a long refactor mid-session.

DeepSeek V4 Flash at $0.139/$0.278 costs 30% less than Luna on input and 4.3x less on output, with a 1M context window. It swaps into an existing Codex CLI or any OpenAI-SDK client by pointing base_url at the endpoint above and sending a Morph API key (setup in our Codex provider configuration guide). The tradeoff is honest. You give up OpenAI's cloud features, managed integrations, and the specific GPT-5.6 weights. You keep uncapped throughput and per-token billing. See the model lineup, full pricing, and the best open-source LLMs for coding.

Frequently Asked Questions