Codex comes with ChatGPT Plus at $20 a month, Pro from $100 and Business at $25 per user ($20 billed yearly), and Free and Go ($8) run GPT-6 Luna in the desktop app. Past a plan's limits you buy credits or run Codex on an API key at API prices.
Claude CodeFable 5.1 (max) (with fallback)70.4 · Coding Agent Index
CodexGPT-6 Astra (max)67.0 · Coding Agent Index
Muse CodeMuse Spark 1.3 (xhigh)64.2 · Coding Agent Index
Grok BuildGrok 4.5 (high)64.1 · Coding Agent Index
Kimi Code CLIKimi K362.6 · Coding Agent Index
OpencodeGemini 3.8 Flash (high)61.2 · Coding Agent Index
Cursor CLIComposer 2.5 Fast38.3 · Coding Agent Index
Demo setups use current model families. The pairings above are the combinations benchmarked by Artificial Analysis; rankings measure performance, not popularity.
Illustrative daily trends · Other totals: last 30 days. Build Hours: this week, model-estimated activity including overlap; not engineer hours.
USED BY ENGINEERING TEAMS AT
How much does Codex cost?
Codex comes with ChatGPT plans and counts against the plan's limits, which ChatGPT Work shares. Per person or user:
ChatGPT plans that include Codex, per person or user. Prices verified October 12, 2026 against learn.chatgpt.com/docs/pricing and OpenAI's Help Center.
Plan
Price
Codex in it
Free and Go
$0 and $8 a month
GPT-6 Luna in the desktop app, for quick tasks, subject to rollout
Plus
$20 a month
Web, CLI, IDE extension and iOS, with GPT-6.1 Sol and GPT-6 Luna
Pro
$100, $200 or $500 a month
More usage and no five-hour limit for now; the top tier adds Ultrafast mode
Business
$25 per user a month, or $20 billed annually
From 2 users, with admin controls and SAML SSO
Business Premium
$125 per user a month, or $100 billed annually
5x a Standard seat's usage and no five-hour limit; mixes with Standard seats
Enterprise and Edu
Priced by sales
SCIM, role-based access, user analytics and audit logs
API key
API prices per token
CLI, SDK and IDE extension, without cloud features such as GitHub code review
What does Codex cost a team of 10?
Business is the team plan, from 2 users. Past the limits, workspaces on flexible pricing buy credits; what a credit costs depends on the plan or agreement, so it isn't in the sum.
A team of 10 on Codex, per month before tax, at the plans table's prices (read October 5, 2026).
What the team buys
Arithmetic
Per month
Business, billed annually
10 × $20
$200
Business, billed monthly
10 × $25
$250
8 Standard and 2 Premium seats, billed annually
8 × $20 + 2 × $100
$360
Business, with credits past the limits
10 × $20, plus credits at your plan's credit price
From $200
Is Codex cheaper than Claude Code?
Per token, yes. At API list prices GPT-6.1 Sol costs half what Claude Opus 5.5 does, for input, cached input and output alike. On a plan, the entry tiers cost the same per month, and Codex Pro has no five-hour limit for now; Anthropic's side is on Claude Code pricing.
What decides a bill is how many tokens each agent spends on the same work, and we can't measure that fairly: our team runs mostly Claude Code. How the two compare on our own pull requests is on Claude Code vs Codex.
How heavy Codex use looks: our numbers
On a plan, Codex reports tokens and limits, not dollars. One engineer's Codex logs, and what Codex wrote in our two repos:
One engineer's Codex logs from June 21 to September 22, 2026 (when they stop), and pull requests merged in our two repos to October 3, 2026.
Measured
Value
Codex sessions
368, of which 188 were subagent threads
Input tokens
6.55 billion, 96.7% of them cached
Reasoning
36% of output tokens
Days the weekly limit reached 100%
7, all in September
Days the five-hour limit reached 100%
2, both in June
Codex's share of merged pull requests
5% from July 5; 9.9% in the last 30 days
Pull requests Codex opened in September
31, merged a median 37 minutes after opening
Codex API pricing per token
With an API key, Codex bills these rates per token and drops cloud features such as GitHub code review. On a plan, usage past the limits takes credits at the last column's rates: 25 credits per dollar of the API price, with no cache-write charge. One GPT-6.1 Sol turn in credits:
GPT-6.1 Sol, priced from the table below
One turn
cached input 200,000 × 2.5 credits per million = 0.5 credits
new input 10,000 × 50 credits per million = 0.5 credits
output 2,000 × 250 credits per million = 0.5 credits
total 1.5 credits
OpenAI API prices per million tokens (standard tier, prompts up to 272K tokens), and the Codex credits the same tokens take. Prices verified October 5, 2026 against developers.openai.com/api/docs/pricing and learn.chatgpt.com/docs/pricing.
Model
Input
Cache write
Cached input
Output
Codex credits (input / cached / output)
GPT-6 Astra
$10
$12.50
$1
$50
250 / 25 / 1,250
GPT-6.1 Sol
$2
$2.50
$0.10
$10
50 / 2.5 / 250
GPT-6 Luna
$0.10
$0.125
$0.01
$0.50
2.5 / 0.25 / 12.5
Codex usage limits by plan
On Plus and Standard Business, OpenAI estimates 5–45 local messages per five hours on GPT-6 Astra, 15–160 on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna. Cloud tasks use more of the allowance, and weekly limits may also apply.
Pro and Business Premium seats have no five-hour limit for now. Enterprise and Edu without flexible pricing get Plus's per-seat limits for most features; with flexible pricing they have no fixed rate limits.
At the limit, a turn already running can finish. After that, Plus and Pro users buy credits, Business, Enterprise and Edu workspaces on flexible pricing buy workspace credits, and anyone can run local chats on an API key at API rates.
Credits are charged per million tokens at a rate for each model, with no separate charge for cache writes. What a credit costs depends on your plan or agreement.
Speed modes spend the allowance faster: Fast mode uses included limits at 2.5x the Standard rate (2x on purchased credits), and Ultrafast, on GPT-6 Astra or GPT-6.1 Sol, at 8x (6x on credits).
What drives a team's Codex bill
OpenAI says model choice, context, reasoning, tool use and caching all affect usage. Four things matter most:
The model. Per token, GPT-6 Astra takes 5 times the credits of GPT-6.1 Sol, and GPT-6 Luna a twentieth of Sol's.
Cached input. It takes a tenth of the credits of new input on GPT-6 Astra and a twentieth on GPT-6.1 Sol, so context Codex has already read costs far less.
Cloud or local. Cloud tasks use more of the allowance than local messages.
Context on every message. OpenAI suggests a smaller AGENTS.md and fewer MCP servers, since each server adds context to every message.
How to see Codex cost per engineer
OpenAI's own tools first:
Each developer: the Codex usage dashboard shows current limits and reset times, and /status shows them in the CLI.
Admins: the Admin Console's Usage and Insights views show activity, credits and tasks by category.
Enterprise and Edu: a Codex analytics dashboard, and the Daily Usage Analytics API for aggregated daily ChatGPT, Codex and Work usage.
Zest: the Codex plugin syncs each session's model and tokens, priced at list rates, per engineer, next to their Claude Code and Cursor sessions. More in Codex usage.
TRY IT ON YOUR TEAM
See what your team ships with AI.
Starter is free (3 seats, no credit card). Pro is $49 a month with 5 seats and Max is $149 a month with 10 seats — priced per workspace, not per engineer, with extra seats at $19 a month.
Free and Go include GPT-6 Luna in the desktop app for quick tasks, subject to rollout. Codex in the CLI, IDE extension and cloud starts with Plus.
What is Codex on ChatGPT Business?+
ChatGPT Business is the team plan that includes Codex, from 2 users, priced per user (see the plans table). It adds admin controls, SAML SSO and larger cloud machines, and workspaces can buy credits past the limits.
Can I see Codex usage per engineer?+
Each developer sees their own limits in the Codex usage dashboard, and admins see activity and credits by category in the Admin Console. Zest shows each engineer's Codex sessions, tokens and list-price cost next to the pull requests they led to.
Winding / Teams
