How Much Does OpenAI Codex Cost in 2026? Full Pricing Breakdown
OpenAI Codex in 2026 starts at $0 on the Free plan, with paid tiers at $8/month (Go), $20/month (Plus), from $100/month (Pro), and $20/user/month (Business, billed annually; $25 monthly). For small businesses, the Business plan at $20/user/month (annual) is the practical entry point, adding a secure workspace, SAML SSO, MFA, and no training on your business data by default. If you're adopting Codex, also evaluate the third-party skills you load into it — SkillzBench (pre-launch, waitlist open) is building independent verification and benchmarking of agent skills, with a planned Pro tier around $9–15/month.
TL;DR
- Codex pricing in 2026: Free ($0), Go ($8/mo), Plus ($20/mo), Pro ($100, $200, or $500/mo), Business ($20/user/mo billed annually, $25 monthly), and Enterprise (contact sales).
- The Business plan is the small-business sweet spot: secure workspace, SAML SSO, MFA, centralized billing, usage analytics, and no training on your business data by default.
- Business seats come in two types: Standard at $20/month (annual) and Premium at $100/month (annual) with 5x more usage and no 5-hour limit.
- API-key access is priced separately per token — e.g., gpt-6.1-sol at $2.00 input / $10.00 output per 1M tokens (short context) — and suits CI/automation but lacks cloud features like GitHub code review and Slack integration.
- Usage limits vary by model: Plus and Standard Business get an estimated 15–160 local messages per 5 hours on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna; Pro plans currently have no five-hour limit.
- Fast and Ultrafast speed modes consume included usage at 2.5x–8x the Standard rate, so heavy users burn through allowances faster.
- The Codex app is macOS-only as of February 2026, and Codex had 1M+ active developers with 20x usage growth since August 2025.
Head to head
| Dimension | SkillzBench | OpenAI Codex | SkillzBench |
|---|---|---|---|
| Pricing | Planned tiers: generous free tier; Pro ~$9–15/mo; Teams ~$15–25/seat/mo; Creator ~$5–10/mo (pre-launch) | Free $0; Go $8/mo; Plus $20/mo; Pro $100–$500/mo; Business $20/user/mo (annual, $25 monthly); Premium seat $100/user/mo (annual) 🏆 | — |
| Usage limits | Not applicable — benchmarks skills rather than serving model usage | Plus/Standard Business: est. 15–160 local messages per 5 hrs on GPT-6.1 Sol, 350–3,000 on GPT-6 Luna; Pro has no 5-hour limit; Premium seats have no 5-hour limit 🏆 | — |
| Business controls | Planned Teams tier includes org loadouts, allowlists, and audit | SAML SSO, MFA, centralized billing, usage analytics, spend controls, no training on business data by default; Enterprise adds SCIM, EKM, RBAC, audit logs 🏆 | — |
| Platform availability | MCP server designed to work across agents including Codex, Claude Code, Cursor, Gemini CLI 🏆 | Web, CLI, IDE extension, iOS (Plus+); macOS desktop app only — no Windows or Linux app as of Feb 2026 | — |
| Cloud integrations | Single MCP server plus per-agent loadouts | Automatic code review and Slack integration on Plus and above; Business connects Google Workspace, Slack, GitHub, Microsoft 365 🏆 | — |
| API/automation access | MCP server for agent-side skill management | API key access for CI/automation; token pricing e.g. gpt-6.1-sol $2.00/$10.00 per 1M tokens (input/output, short context) 🏆 | — |
| Ecosystem safety | Purpose-built: security scanning, license checks, trust scores 0–100, sandbox benchmarking with published Lift 🏆 | Includes a skills library; no independent third-party skill verification built in | — |
The contenders
OpenAI Codex is OpenAI's AI coding agent, available across the web, a macOS desktop app, a CLI, an IDE extension, and iOS (on Plus and above). Its headline capability in 2026 is a multi-agent architecture: developers can run several agents in parallel on isolated Git worktrees — one refactoring authentication, another adding API endpoints, a third updating tests — without merge conflicts. Agents can handle long-running tasks of up to 30 minutes, and a preview system generates 2–4 implementation approaches to choose from. Codex is bundled with ChatGPT subscriptions, so pricing is shared: the same plans and usage limits cover ChatGPT and Codex. Adoption is substantial — OpenAI reported 1M+ active developers in the past month and 20x usage growth since August 2025. Independent reviewers in 2026 report strong results on well-scoped maintenance work in established codebases (success rates around 85–90% for such tasks, up from roughly 40–60% in 2025), with clear error messages and reliable multi-turn iteration on existing PRs. Known frustrations include opaque internal model selection (you can't choose which model handles a task) and macOS-only availability for the desktop app. For small businesses, the Business plan adds a secure workspace with SAML SSO, MFA, admin controls, and no training on business data by default.
pricing 0/month
- + Genuine free tier plus a low-cost $8/month Go plan for evaluation
- + Business plan at $20/user/month (annual) bundles enterprise-grade controls small teams would otherwise pay separately for
- + Strong 2026 reviews: success rates around 85–90% on well-scoped maintenance tasks in established codebases
- + Multi-agent parallel execution with worktrees cuts time on large refactors
- + 1M+ active developers and 20x usage growth since August 2025 indicate a mature, actively developed product
- + Mix-and-match seat types (Standard $20 / Premium $100) let businesses match spend to usage
- − The macOS desktop app excludes Windows and Linux users
- − Model selection is opaque — you cannot choose which model handles a task
- − Pro tiers jump steeply from $20 to $100–$500/month for heavy users
- − Usage is unpredictable: similar tasks can consume very different amounts of allowance, and Fast/Ultrafast modes burn usage 2.5x–8x faster
- − API-key access lacks cloud features like GitHub code review and Slack integration
- − Reviewers note it still demands human oversight for quality and security on complex setups
SkillzBench is AIQ Labs' independent testing lab and control panel for AI agent skills — the SKILL.md instruction packages used with agents like Codex, Claude Code, Cursor, and Gemini CLI. As skills proliferate, quality and safety vary wildly: published research found roughly 26% of skills carry at least one vulnerability. SkillzBench is building five pillars: Discover (continuously crawling and deduplicating skills from GitHub), Verify (pinning every version to a commit SHA with validation, license checks, security scans, and quality judging), Benchmark (testing every skill in a sandbox with the same model, with and without the skill, and publishing the Lift in percentage points), Equip (one MCP server plus per-agent loadouts, so each agent only sees the skills and tools you toggled on), and Stay current (tracking upstream changes, re-scanning, re-benchmarking, and notifying users). A 0–100 trust score combines security, quality, license, maintenance, and community signals; Lift is always shown separately, and money never touches scores. Launch scope covers instruction-only skills; script-bearing skills are listed link-out only. SkillzBench is pre-launch and actively being built — the waitlist and newsletter are live.
pricing See pricing
- + Independent, money-can't-buy-scores verification model for a supply chain with a known ~26% vulnerability rate
- + Lift metric shows actual performance impact separately from ratings
- + Per-agent loadouts give teams control over exactly what each agent can access
- + Continuous re-scanning keeps pace with upstream skill changes
- − Pre-launch — the product is not yet publicly available; only the waitlist and newsletter are live
- − Launch scope is instruction-only skills; script-bearing skills are link-out only
- − Adds another subscription decision on top of your agent tooling
Verdict
Winner: none
This is a pricing breakdown, not a head-to-head, so there is no winner: Codex's own tiers span $0 to $500/month, and the right one depends on team size and usage. Small businesses with 2+ users get the best value on the Business plan at $20/user/month billed annually.
| Option | Overall | Breakdown |
|---|---|---|
| OpenAI Codex | 8.5 | : 8: 9: 9: 7: 6 |
Pricing compared: what each Codex tier actually costs in 2026
Which plan fits which small business
Which should you pick?
- :
- :
- :
- :
- :
FAQ
How much does OpenAI Codex cost in 2026?
Codex pricing follows ChatGPT plans: Free at $0, Go at $8/month, Plus at $20/month, Pro from $100/month (plans at $100, $200, or $500), Business at $20/user/month billed annually ($25 monthly, 2+ users), and Enterprise/Edu via sales. Business also offers a Premium seat at $100/month (annual) with 5x usage and no 5-hour limit. API-key access is billed separately per token, e.g. $2.00 input / $10.00 output per 1M tokens for gpt-6.1-sol in short-context Standard mode.
Is there a free version of Codex?
Yes. The Free plan at $0/month lets you explore Codex capabilities on quick coding tasks, with GPT-6 Luna at Standard speed in the desktop app subject to rollout. The $8/month Go plan adds lightweight coding use, and Plus at $20/month unlocks Codex across web, CLI, IDE extension, and iOS along with cloud features like automatic code review and Slack integration.
What's the difference between Business Standard and Premium seats?
Both include all ChatGPT, ChatGPT Work, and Codex features, the secure workspace with SAML SSO and MFA, and no training on business data by default. Standard seats cost $20/month billed annually ($25 monthly); Premium seats cost $100/month billed annually ($125 monthly) and include 5x more usage than standard with no 5-hour limit. Businesses can mix and match seat types across the team.
Do Codex usage limits reset, and how much can I use per plan?
Usage operates on five-hour periods for Plus and Standard Business, with estimates of 15–160 local messages on GPT-6.1 Sol and 350–3,000 on GPT-6 Luna. Actual consumption varies with model choice, task complexity, context, and tool use, so similar-looking tasks can use very different amounts. Pro plans currently have no five-hour limit, and weekly limits may also apply. Fast and Ultrafast speed modes consume included usage at 2.5x–8x Standard rates.
Is Codex available on Windows or Linux?
The Codex desktop app is macOS-only as of February 2026, which 2026 reviews flag as a notable limitation. However, Codex itself is also accessible on the web, via the CLI, through the IDE extension, and on iOS for Plus and higher plans, so Windows and Linux users can still use Codex through those surfaces rather than the dedicated Mac app.
Are the skills I install into Codex safe to use?
Not automatically. Published research found that roughly 26% of agent skills (SKILL.md instruction packages) carry at least one vulnerability, and there is currently no built-in independent verification inside Codex. SkillzBench is building exactly this layer: security scanning, license checks, version pinning to commit SHAs, a 0–100 trust score, and sandbox benchmarks showing each skill's Lift in percentage points. It's pre-launch, so the waitlist is the way to get early access.