Codex Pricing: GPT-6.1 Sol, GPT-6 Astra, Plans, Credits, and API Costs

Bottom line: how Codex pricing works

Codex pricing is organized around three distinct financial tracks rather than a single standalone software license:

1. Included with ChatGPT plans: Individual developers accessing Codex through ChatGPT Plus ($20/mo, expanded Codex usage) or Pro (from $100/mo, maximum Codex tasks) receive included access to coding tools, code interpretation sandboxes, and conversational engineering assistance. Free and Go include limited Codex access. Teams deploying ChatGPT Business ($25/user/mo monthly or $20 annual) receive all Codex features, centralized administrative management, and the option to buy credits for more access.
2. Autonomous cloud coding tasks and credits: When executing long-running, multi-file autonomous tasks in containerized cloud sandboxes, Codex draws from included monthly allowances or supplemental purchased credit packs. Once included task allowances are consumed, administrators can enable on-demand billing bounded by hard spending caps.
3. Separately billed OpenAI developer API: Programmatic API calls made with developer API keys on platform.openai.com (such as integrating OpenAI models into custom CI/CD pipelines, IDE plugins, or internal tools) are billed strictly on a metered per-token basis. ChatGPT subscriptions and Codex workspace fees never cover API token invoices.

Legacy Codex-only Business standalone seats (from prior to the 2026 workspace consolidation) remain active only for organizations established prior to June 2026; all new teams subscribe through ChatGPT Business or Enterprise plans.

Access routes and tier comparison matrix

The table below outlines the pricing, included coding allowances, model capabilities, and spending controls across the primary ways developers access Codex.

Understanding Codex compute: interactive editing vs cloud tasks

Codex compute consumption falls into two distinct operational paradigms, each with different cost implications:

1. Interactive Canvas and conversational assistance

When a developer uses Codex inside ChatGPT to refactor a function, explain an unfamiliar code snippet, or generate unit tests, compute is treated as standard chat inference.

• Covered completely by your monthly Plus ($20), Pro (from $100), or Business ($25/seat) subscription.
• Subject only to your plan's standard usage limits for the selected model.
• Does not consume specialized autonomous task credits.

2. Autonomous cloud tasks and execution sandboxes

When Codex is tasked with resolving a complex engineering issue—such as cloning a repository, installing dependencies, editing multiple interrelated files, running unit tests, and iteratively debugging build failures—it executes inside a dedicated, containerized cloud sandbox.

• Task allowance consumption: Cloud agent workflows draw down your monthly included task credits.
• Compute intensity: A complex task that runs for 10 minutes, executes multiple terminal commands, and iterates through five failed test runs consumes significantly more compute than a standard chat response.
• Supplemental credit packs: When a team exhausts its included monthly cloud task allowance, workspace administrators can purchase supplemental credit packs to ensure deployment pipelines remain unblocked.

Codex Billing and Usage Flow

Developer triggers coding workflow

• Is this interactive chat / Canvas?
• • YES: Covered by ChatGPT Subscription ($8–$500/mo)
  • NO (Autonomous cloud agent task): Are included workspace task credits available?
  • • YES: Deduct from included plan allowance
    • NO: Is On-Demand / Supplemental Credits enabled?
    • • YES: Within Hard Spend Cap?
      • • YES: Execute in cloud sandbox
        • NO: Pause task until admin approval
      • NO: Task halted: prompt user to upgrade or add credits

Spend caps and administrative governance

In team environments, unmonitored autonomous agent loops can inadvertently trigger escalating cloud bills. OpenAI provides structured administrative guardrails within ChatGPT Business and Enterprise consoles:

• Hard Spend Caps: Workspace owners can set a strict monthly monetary limit (e.g., $100 or $500 per team) on supplemental task credits. Once this ceiling is reached, Codex will not initiate new paid autonomous agent runs without manual admin authorization.
• Soft Limit Notifications: Email alerts are triggered when monthly credit consumption reaches 75% and 90% of the defined threshold.
• Seat-level permissioning: Administrators can restrict autonomous cloud task execution to senior engineers while granting junior developers interactive Canvas access only.

The subscription vs API billing boundary

A frequent question from engineering organizations is whether paying for ChatGPT Plus or Business covers the API tokens needed for developer tools like Cursor, Cline, or custom CLI scripts.

• The strict boundary: Subscriptions at chatgpt.com grant access to graphical web interfaces, Canvas, and official mobile apps. They never include API token credits.
• The developer platform: In order to generate an sk-... API key for external IDE integrations or custom backend services, you must create an account at platform.openai.com, add a payment card, and pre-purchase API balance tiers.
• Invoicing for API usage occurs separately based on exact input and output token counts per model (e.g., $2.00 per million input tokens for GPT-6.1 Sol and GPT-6 Sol, with cached input at $0.10 on GPT-6.1 Sol and $0.20 on GPT-6 Sol).

Unit economics: Codex vs Cursor vs GitHub Copilot vs Claude Code

When budgeting for developer productivity tools, organizations should compare the total cost of ownership across major coding assistants:

Buyer recommendation checklist

To choose the right Codex access path for your workflow:

1. Choose ChatGPT Plus ($20/mo) if you are an individual software developer who wants interactive code refactoring, algorithm design, and code explanation via Canvas alongside general AI research capabilities.
2. Choose ChatGPT Pro (from $100/mo) if you need maximum Codex tasks and Pro reasoning powered by GPT-6 Astra for complex architectural refactoring, algorithmic proofs, or heavy data engineering tasks that exceed Plus usage limits.
3. Choose ChatGPT Business ($25/seat/mo, or $20 annual) if you are an engineering team lead looking to provide developers with collaborative coding tools, shared custom GPTs, centralized billing, and no training on business data by default.
4. Use the OpenAI Platform API if your primary goal is building custom internal developer automation, CI/CD bots, or connecting models to third-party editors like Cursor or Claude Code.