Subscription vs API

Weigh an AI subscription against pay-as-you-go API at your usage level. Subscriptions bill per time window rather than per token, so no exact break-even exists: results are given as ranges, with the cache sensitivity and the date each plan was checked.

Plan prices and quotas are maintained by hand and have not been checked with each provider individually. API prices are read from the catalog, so changing the price basis can reverse the conclusion. Quota rules change frequently; the official page takes precedence.

Prompts / day

Input tokens / model call

Output tokens / model call

Cache hit rate 70%

Model calls per prompt: 8

Chat is about 1. An agent plans, searches, calls tools and summarises, so one prompt often triggers 5 to 30 calls. That is how a model with a low per-token price can still be expensive per task.

API price basis

First-party list price, the like-for-like alternative to a first-party subscription.

Scenario presets

≈ 4,800 model calls / month at this usage.

Cost ranges come from cache-hit uncertainty (±15pp). Capacity ranges come from reading each window quota conservatively vs optimistically. The break-even is a region, not a point, because subscription quotas are time windows that cannot be reduced to a fixed token count.

Rolling windows shorter than a day are normalised against an assumed 176 active hours per month (8h × 22 days); day, week and month caps are calendar windows and are not discounted.

Plans without a computable quota

These publish no absolute quota, or bill in a unit that cannot be mapped to model calls. They are listed here rather than given an invented number.

OpenAI FreeFreeModel and feature limits are dynamic and no absolute monthly or message cap is published.
OpenAI Go$8.00Higher limits than Free are published without an absolute message or monthly allowance.
OpenAI Plus$20.00Advanced-model and feature limits are dynamic; the pricing page gives no absolute API-call equivalent.
OpenAI Pro $100$100.00The 5× usage statement is relative to Plus and cannot be converted to an absolute API-call allowance.
OpenAI Pro $200$200.00The 20× usage statement has no absolute API-call equivalent; new subscriptions and upgrades are temporarily paused.
OpenAI Business Standard$25.00Business usage is subject to service guardrails and has no absolute published API-call allowance.
OpenAI Business Premium$125.00Premium usage is relative to the Standard seat and has no absolute published API-call allowance.
Anthropic FreeFreeUsage depends on the model, conversation and feature; no absolute cap is published.
Anthropic Pro$20.00Usage credits and weekly limits depend on task and model; no absolute API-call allowance is published.
Anthropic Max 5x$100.00The 5× statement is relative to Pro and task-dependent, not an absolute API-call allowance.
Anthropic Max 20x$200.00The 20× statement is relative to Pro and task-dependent, not an absolute API-call allowance.
Anthropic Team$25.00Per-seat usage is not published as an absolute API-call allowance.
Google AI Plus$4.99The plan states a relative 2× usage limit; compute-based limits cannot be converted to API calls.
Google AI Pro$19.99The plan states a relative 4× usage limit; compute-based limits cannot be converted to API calls.
Google AI Ultra 5x$99.99Usage is measured as compute and varies with the prompt, features and conversation; no API-call cap is published.
Google AI Ultra 20x$199.99Usage is measured as compute and varies with the prompt, features and conversation; no API-call cap is published.
xAI FreeFreeRate limits are not published as absolute model-call numbers.
xAI SuperGrok$30.00Higher rate limits are not published as absolute model-call numbers.
xAI SuperGrok Plus$100.00Higher rate limits are not published as absolute model-call numbers.
Perplexity FreeFreePro-search and feature limits are not equivalent to a fixed model-call allowance.
Perplexity Pro$20.00Search and research usage bundles retrieval, tools and generation, so it cannot map to API model calls.
Perplexity Max$200.00Monthly AI credits vary in cost by research feature and model, so they cannot map to API model calls.
Mistral FreeFreeVibe feature limits and Studio credits do not specify a fixed model-call allowance.
Mistral Pro$14.99Vibe feature limits and included Studio API credits do not specify a fixed model-call allowance.
Mistral Team$24.99Per-seat Vibe use and included credits do not specify a fixed model-call allowance.
GitHub FreeFreeChat, agent and completion limits are product-specific and are not API model-call limits.
GitHub Pro$10.00Included GitHub AI Credits have feature- and model-dependent costs, not a fixed API-call conversion.
GitHub Pro+$39.00Included GitHub AI Credits have feature- and model-dependent costs, not a fixed API-call conversion.
GitHub Max$100.00The $100 monthly AI Credit allocation has feature- and model-dependent costs, not a fixed API-call conversion.
GitHub Business$19.00Included GitHub AI Credits have feature- and model-dependent costs, not a fixed API-call conversion.
GitHub Enterprise$39.00Included GitHub AI Credits have feature- and model-dependent costs, not a fixed API-call conversion.
Cursor HobbyFreeAgent-request limits are not published as absolute model-call numbers.
Cursor Pro$20.00Agent limits and usage-based features do not specify a fixed model-call allowance.
Cursor Pro+$60.00Agent limits and usage-based features do not specify a fixed model-call allowance.
Cursor Ultra$200.00Agent limits and usage-based features do not specify a fixed model-call allowance.
Cursor Teams Standard$40.00Team use and usage-based features do not specify a fixed model-call allowance.