OpenAI
OpenAI Codex
OpenAI's local and cloud coding agent for terminal work, desktop multi-agent workflows, code review, and hosted tasks.
Attributed output
GitHub public-attribution count, 2026-06-09
5.1M PRsAttributed outputGitHub public-attribution count, 2026-06-09Commits and PRs on public GitHub provably produced by this agent. Exact if it is visible through its own bot or workflow, an estimate if measured through attribution it declares (these can swing 30 to 40 percent between measurement runs at multi-million scale), and a floor either way.#4
Index rank
of 34 tracked agents
#4Index rankof 34 tracked agents$20/mo Plus
Entry price
Plan limits plus token-based credits
$20/mo PlusEntry pricePlan limits plus token-based credits10,345,876
npm demand/wk
Adjusted from raw npm channel
10,345,876npm demand/wkAdjusted from raw npm channelWeekly package downloads as an install-demand floor; adjusted when the vendor's main install channel is no longer npm.318,191,030
npm installs to date
117,184,916 excluding the CI spike
318,191,030npm installs to date117,184,916 excluding the CI spikeTotal npm downloads of this agent's package(s) summed across every tracked week since it was first published. A floor, like the weekly figure. Where a one-week automated/CI burst inflated the total, the baseline-only figure is shown too.Activity Over Time
Weekly attributed output on public GitHub. Output visible through the agent's own bot or workflow is exact; output measured through declared attribution is an estimate. Window: 2025-02-03 to 2026-06-08.
npm Installs Over Time
Weekly npm downloads since 2025-04-14, the install-demand trajectory. Toggle Ex-CI spike to remove the one-week automated burst.
Market
Developer output
Public pull requests from Codex cloud
Counted from the pull requests Codex's cloud service opens. That cloud path is well-covered (about 86% merge), but the match is inferred from public pull-request signals rather than an exact bot identity, so treat the count as an estimate, not an exact figure. Codex's local command-line work runs under the developer's own identity, so that output is largely invisible to public attribution and not in this count, making the total a floor. An earlier signal was retired in late 2025 and replaced, so the series stays continuous.
Last full week (2026-06-08): 125K agent-marked PRs (-9.7% week over week), search estimate.
Distinct users: 4,258 humans opened the 12,600 most recently sampled PRs, about 3 PRs per person.
Merge rate: 87.1% of its pull requests get merged (963,705 of 1,105,827 over 8 settled weeks, 2026-04-13 to 2026-06-07). The window ends about ten days back so recently-opened PRs have had time to merge.
GitHub review activity: 2.2% of its merged PRs had at least one GitHub review record before merging; the other 97.8% had none. This is a broad weekly signal and can include bot reviews; the human-review read below comes from the enriched PR sample.
In significant repos (company-owned or 10+ stars), 38.7% of its merged PRs got a human review. Across all sampled repos, the rate is 24.2%. Human review means a non-author, non-bot GitHub reviewer in the PR-lifecycle sample.
Median time to merge: 11 min. The unreviewed median is 0.1h. The human-reviewed median is 1.55h. From a 120-PR sample merged 2026-05-07 to 2026-06-05.
Test-inclusion: 40.8% of its sampled merged PRs touch a test file. This is a rough read on production readiness. 5 large PRs hit GitHub's 100-file path read cap, so treat this as a floor.
Flow share: 1.1% of the 7998 most recent public PRs sampled (86 PRs, 8 windows, last 2026-06-17).
Snapshot 2026-06-09, confidence Medium, tracked since 2026-06-09, +1.2% over 6 days. Read the methodology.
- Index rank
- #4 of 34
- Demand share
- 40% of tracked npm demand floor
- Business signal
- 5M+ weekly Weekly Codex users
- GitHub share
- 8% of tracked stars
- npm channel
- 9.8M raw downloads/week (npm primary)
Best fit
Direct competitors
- OpenAI says non-developers are about 20% of Codex users and are growing more than 3x as fast as developers.
- Codex revenue is not disclosed separately from ChatGPT, Business, Enterprise, or API consumption.
npm channel caveat: npm primaryExpandHide
Codex has a steady npm install channel (~6M/week) plus a one-week spike in early May 2026.
Daily npm installs sit around 0.8-1M (~6M/week) before and after the spike, so distribution did not stop. For one week (Apr 30 to May 6, 2026) daily downloads jumped 30-50x to 13-46M, then fell straight back to baseline. That sudden-then-gone shape is characteristic of automated or CI traffic around a release (a transitive dependency or pinned action), not a durable change in active users, so the demand figure here uses the trailing 4-week average and excludes the spike.
- Demand floor
- 10.3M / wk
- Latest npm week
- 9.8M
- Peak npm week
- 130.5M
- 4-week average
- 10.3M
Trailing 4-week average; release spike excluded from demand floor
Complete Mon-Sun week when cron-backed
Observed public npm range
As of 2026-06-14
All sourced signals (2)ExpandHide
| Signal | Value | Type | Confidence |
|---|---|---|---|
| GitHub stars | 89,162 | Public metric | High |
| npm weekly downloads | 13,450,673 | Public metric | High |
Where OpenAI Codex Works
Share of OpenAI Codex's own classified sampled work in each segment. Full breakdowns: Languages, Frameworks, Industries.
Pricing And Economics
ExpandHide
Codex is bundled in ChatGPT plans; Pro is now split into $100 5x and $200 20x tiers, with token-based Codex credits.
- Entry price
- $20/mo Plus
- Top self-serve price
- $200/mo
- Pricing unit
- Plan limits plus token-based credits
- Token transparency
- Token rate card
- Included usage
- Plus has published 5-hour Codex limits by model; Pro $100 unlocks 5x Plus usage and Pro $200 unlocks 20x Plus usage.
- Overage path
- Plus/Pro can buy credits; Business Codex has no fixed seat fee and bills pay-as-you-go by usage.
Retail Token Equivalent
Codex credits are now charged per 1M input, cached input, and output tokens. GPT-5.5 is 125 / 12.5 / 750 credits per 1M tokens; GPT-5.3-Codex is 43.75 / 4.375 / 350 credits per 1M tokens. A typical GPT-5.5 Codex task is published at roughly 5-45 credits.
| Plan | Price | Allowance | Unit | Token / Overage Notes |
|---|---|---|---|---|
| ChatGPT Plus | $20/mo | Expanded Codex usage; example Plus limits include GPT-5.3-Codex at 30-150 local messages, 10-60 cloud tasks, and 20-50 code reviews per 5h. | 5-hour messages/tasks | Actual usage is now token-metered through Codex credits. Buy additional credits or run local tasks with an API key at API rates. |
| ChatGPT Pro 5x | $100/mo | 5x higher usage than Plus for Pro capabilities and Codex. | relative plan multiplier | Credit burn depends on token mix, model, speed mode, and task size. |
| ChatGPT Pro 20x | $200/mo | 20x higher usage than Plus for the highest self-serve Pro tier. | relative plan multiplier | Credit burn depends on token mix, model, speed mode, and task size. |
| Business Codex | Usage pricing | No fixed seat fee; usage-based development workspace. | credits/tokens | Pay as you go based on token-priced Codex credits. |
Unit Economics (Measured)
Mean API-retail cost per completed task on the Artificial Analysis coding agent index (358 tasks across 3 suites, pass@1 over 3 runs, snapshot 2026-06-09). These are pay-per-token API costs, not subscription effective costs.
| Harness | AA index | Cost / task | Tokens / task | Time / task |
|---|---|---|---|---|
| Codex, GPT-5.5 (xhigh) | 65.3 | $4.33 | 4.8M | 9m |
| Codex, GPT-5.5 (medium) | 60.4 | $2.21 | 2.8M | 7m |
| Codex, GPT-5.4 (medium) | 53.6 | $2.09 | 2.5M | 7m |
Break-even vs API retail on the top harness (Codex, GPT-5.5 (xhigh), $4.33/task): ChatGPT Plus at $20/mo covers about 5 tasks; ChatGPT Pro 5x at $100/mo covers about 23 tasks; ChatGPT Pro 20x at $200/mo covers about 46 tasks. Assumes the Artificial Analysis task mix.
Included in ChatGPT Free, Go, Plus, Pro, Business, and Enterprise plans with plan-specific limits; Business Codex is pay-as-you-go.
- ChatGPT Pro advertises 5x or 20x more Codex usage than Plus.
- OpenAI moved Codex pricing to token-based credit rates in April 2026.
- API-key use follows OpenAI API pricing rather than bundled ChatGPT subscription limits.
Company And Funding
ExpandHide
- Ownership
- Private
- Founded
- 2015
- Headquarters
- San Francisco, CA
- Employees
- ~9,300
- Total raised
- ~$180B
- Most recent round
- Series G, $122B (Mar 2026)
- Round leads
- SoftBank, Andreessen Horowitz, Amazon, Nvidia
OpenAI operates as a capped-profit company controlled by the OpenAI nonprofit. Codex is an OpenAI product, not a separately financed entity.
Parent-company total.
Notable investors
Enterprise And Security
ExpandHide
- SSO / SCIM
- Yes
- Deployment
- Cloud-only
- Trains on your code
- Not trained on your code
- Zero data retention
- Enterprise tier
Compliance
Admin controls
OpenAI does not train on Business, Enterprise, or API data by default. Zero data retention is available for eligible endpoints with a qualifying use case.
Product
Providers
Model families
Capabilities
Tradeoffs
- Fast-moving model availability
- Pricing differs across ChatGPT plan, Business Codex, and API-key modes