Pricing, plans, and API prices

Pricing AI CLI

AI CLIs can be paid for in two ways: through a fixed subscription where usage is included up to a limit, or through an API key where you pay for tokens, tool calls, and model choice.

Quick decision

Need	Typical choice
You want to learn the tool	Start with a free tier or small personal subscription.
You use the CLI several times a week	Choose a plan with included CLI usage.
You want automation, CI, or scripts	Use an API key with spend limits.
You work in a team	Use a team or business plan with central billing, SSO, and admin controls.
You work in large repositories	Set budgets, use cheaper models for routine work, and keep context small.

What Drives Cost?

Driver	Why it matters	Practical control
Repository size	The agent may read many files and send more context.	Start in the specific app folder, not the whole workspace.
Instruction files	Long `AGENTS.md`, `CLAUDE.md`, and `GEMINI.md` files are often sent as context.	Keep root files short and put details close to the subproject.
MCP and tools	Each external tool can add descriptions, output, and extra calls.	Enable MCP servers only when the task needs them.
Model choice	Stronger models cost more and often use more reasoning.	Use mini/Haiku/Flash for routine work and stronger models for hard tasks.
Output length	Long reports and generated code cost output tokens.	Ask for concrete findings, patches, and short summaries.
Automation	Loops, cron, and CI can spend without you watching.	Use separate API keys, budgets, rate limits, and logs.

Simple API rule:

price = input tokens + cached input tokens + output tokens + tool calls

For CLI subscriptions, the practical rule is:

usage = number of tasks * task size * model choice * extra tools/subagents

OpenAI Codex and OpenAI API

Codex can use ChatGPT login or an OpenAI API key. ChatGPT plans include Codex access across web, CLI, IDE, and sometimes cloud features. API key usage is best for automation and CI, but cloud features such as GitHub code review and Slack are not included.

Plan	Price	Codex relevance
Free	$0/mo.	Quick test tasks and learning.
Go	$8/mo.	Lightweight coding tasks.
Plus	$20/mo.	Codex in web, CLI, IDE, and iOS, newer models, and credits.
Pro	From $100/mo.	5x or 20x higher limits than Plus and priority processing.
Business	$20/user/mo. annual or $25/mo.	Team workspace, admin controls, larger VMs, and credits.
Enterprise/Edu	Contact sales	Enterprise controls, audit logs, retention, and data residency.
API key	Token-based	Local Codex tasks in CLI/SDK/IDE, but no cloud code review.

OpenAI API prices

Model	Input / 1M tokens	Cached input / 1M tokens	Output / 1M tokens
GPT-5.6 Sol	$5.00	$0.50	$30.00
GPT-5.6 Terra	$2.50	$0.25	$15.00
GPT-5.6 Luna	$1.00	$0.10	$6.00
GPT-5.5	$5.00	$0.50	$30.00
GPT-5.4	$2.50	$0.25	$15.00
GPT-5.4 mini	$0.75	$0.075	$4.50
GPT-5.4 nano	$0.20	$0.02	$1.25
GPT-5.3-Codex	$1.75	—	$14.00

GPT-5.6 Sol is OpenAI’s newest flagship model, with Terra and Luna as the mid-size and small 5.6 variants. Data residency and regional processing can add a 10% surcharge for models released after March 5, 2026.

Codex credits

Codex credits let you continue after included limits. For new and existing Business customers and new Enterprise customers, OpenAI showed this token-based credit card:

Model	Input / 1M tokens	Cached input / 1M tokens	Output / 1M tokens
GPT-5.6 Sol	125 credits	12.5 credits	750 credits
GPT-5.6 Terra	62.5 credits	6.25 credits	375 credits
GPT-5.6 Luna	25 credits	2.5 credits	150 credits
GPT-5.5	125 credits	12.5 credits	750 credits
GPT-5.4	62.5 credits	6.25 credits	375 credits
GPT-5.4 mini	18.75 credits	1.875 credits	113 credits

GPT-5.3-Codex-Spark is a research preview available to ChatGPT Pro users only.

Codex Included Limits

OpenAI gives limits as ranges because they depend on task size and model. As a practical rule, large repositories, long prompts, and subagents use the allowance faster.

Plan	Example local messages per 5 hours
Plus/Business	GPT-5.6 Terra: 20-110
Pro 5x	GPT-5.6 Terra: 100-550
Pro 20x	GPT-5.6 Terra: 400-2,200

Cloud tasks and local messages can share the same five-hour window, and weekly limits can also apply.

Anthropic Claude Code and Claude API

Claude Code can be used through Claude subscriptions or API token consumption. Claude Code shows token and plan usage with /usage. Pro and Max subscribers have usage included in their plan, while API users pay per token. Subscriptions give access to Anthropic’s current model families: Fable, Opus, Sonnet, and Haiku.

Plan	Price	Claude Code relevance
Free	$0	Try Claude in the app.
Pro	$17/mo. annual or $20/mo. monthly	More usage, Claude Code, Claude Cowork, and Projects.
Max	From $100/mo.	5x or 20x more usage than Pro and higher output limits.
Team standard	$20/seat/mo. annual or $25/mo.	Team features, Claude Code, SSO, and admin.
Team premium	$100/seat/mo. annual or $125/mo.	5x more usage than standard seats.
Enterprise	$20/seat + API-rate usage	Enterprise controls, spend controls, audit logs, and retention.
API	Token-based	Pay by model, input, output, cache, and tools.

Anthropic describes typical enterprise Claude Code costs as around $13 per developer per active day and $150-250 per developer per month, with large variation based on model, repository size, and usage pattern.

Claude API prices

Model	Input / 1M tokens	Output / 1M tokens	Cache write / 1M	Cache read / 1M
Fable 5	$10	$50	$12.50	$1.00
Opus 4.8	$5	$25	$6.25	$0.50
Sonnet 5, intro through Aug 31, 2026	$2	$10	$2.50	$0.20
Sonnet 5, from Sep 1, 2026	$3	$15	$3.75	$0.30
Sonnet 4.6	$3	$15	$3.75	$0.30
Haiku 4.5	$1	$5	$1.25	$0.10

Opus 4.7 and 4.6 cost the same as Opus 4.8. Cache write prices are for the 5-minute cache; the 1-hour cache costs 2x the input price. Note that Fable 5, Sonnet 5, and Opus 4.7+ use a newer tokenizer that produces roughly 30% more tokens for the same text than Sonnet 4.6 and earlier, so don’t compare on per-token price alone.

US-only inference costs 1.1x on all token prices. Claude web search costs $10 per 1,000 searches. Code execution is free when used with web search/web fetch; otherwise it includes 1,550 free hours per month per organization, then $0.05 per hour per container.

Gemini CLI, Google AI plans, and Gemini API

Gemini CLI has several auth paths: Google login, Gemini API key, and Vertex AI. The right pricing model depends on the login method.

Auth or plan	Price or model	CLI quota in Gemini CLI docs
Google account, Gemini Code Assist individual	Free	1,000 requests per user per day
Google AI Pro	$19.99/mo.	1,500 requests per user per day
Google AI Ultra	$249.99/mo.	2,000 requests per user per day
Gemini API key, unpaid	Free tier	250 requests per user per day, Flash only
Gemini API key, paid	Pay as you go	Varies by pricing tier and token usage
Vertex AI	Pay as you go	Varies by Google Cloud quota and model
Workspace Code Assist Standard	Seat/license	1,500 requests per user per day
Workspace Code Assist Enterprise	Seat/license	2,000 requests per user per day

Google AI Plus exists as a consumer subscription, but the Gemini CLI documentation says tiers that are not listed, including Google AI Plus, are not supported as a CLI paid tier.

Gemini API prices

Model and tier	Input / 1M tokens	Output / 1M tokens
Gemini 3.5 Flash, Standard	$1.50	$9.00
Gemini 3.5 Flash, Batch	$0.75	$4.50
Gemini 3.1 Pro, Standard, under 200k prompt	$2.00	$12.00
Gemini 3.1 Pro, Standard, over 200k prompt	$4.00	$18.00
Gemini 3.1 Pro, Batch, under 200k prompt	$1.00	$6.00
Gemini 3.1 Flash-Lite, Standard text/image/video	$0.25	$1.50
Gemini 3.1 Flash-Lite, Batch text/image/video	$0.125	$0.75

Gemini 3.5 Flash is the latest Flash model and has replaced Gemini 3 Flash as the default choice for fast tasks.

Grounding with Google Search for Gemini 3 models includes 5,000 prompts per month, then $14 per 1,000 search queries. Batch API can reduce token prices, but works best for tasks that can wait.

Practical Budget Choices

Profile	Starting budget	Why
New user	Free tier or one personal plan for a month	You learn workflows before wiring in API keys.
Solo developer	Codex Plus/Pro, Claude Pro/Max, or Google AI Pro by ecosystem	Fixed monthly price is easier than token accounting.
Automation	API key with a low spend limit	Scripts can otherwise spend heavily without you seeing it.
Team	Business/Team/Enterprise	Central billing, SSO, audit, retention, and spend controls matter more than the lowest token price.
Large repository	Subscription plus API fallback	Use the subscription for manual development and API keys for measured automation.

USD and Local Checkout

The English page shows USD because that is how most official pricing pages publish base prices. Treat these as pre-tax references and check checkout for:

VAT or local taxes
currency conversion
local price changes
enterprise discounts
promotions and temporary usage boosts

Avoid expensive surprises

Start with a subscription

For learning and manual use, fixed plans are easier to reason about than API usage.

Set spend limits

API keys should always have budgets, rate limits, and separate keys for test, CI, and production.

Keep context small

Large instruction files, many MCP servers, and broad prompts consume more tokens.

Use a cheaper model first

Use mini, Haiku, or Flash for routine tasks and save the strongest models for hard changes.

Sources

Last checked: July 13, 2026.