InputsUSD · US web-plan prices
Ecosystem

API side uses the vendor's published per-token list prices with prompt caching modeled (chat apps resend conversation history every turn — set cache to 0 for a stack without caching). Thinking models bill hidden reasoning tokens at the output rate — that's real spend in both worlds.

VerdictChatGPT · 30 msgs/day
Same usage at API rates
$26.12/mo
GPT-5.6 Sol · 900 requests/mo

Best value: ChatGPT Plus — this volume would bill ×1.3 its price at API rates.

  • ChatGPT Go$8.00/mo break-even 9/day plan saves $18.12/mo
  • ChatGPT Plus$20.00/mo break-even 23/day plan saves $6.12/mo
  • ChatGPT Pro (5×)$100/mo break-even 110/day API saves $73.88/mo
  • ChatGPT Pro (20×)$200/mo break-even 230/day API saves $174/mo

Plans cap usage in rolling windows and vendors don't publish those caps in tokens — verdicts compare the price of this volume, not entitlements. The higher the multiple, the more likely a cap cuts in first.

Embed this calculator ↓
Start from your profile

Six usage profiles

Click one to load it. Verdicts are computed from the current plan sheet and API snapshot — they move when a vendor reprices.

All plans

Your usage against every plan

The scenario configured above, priced against all 11 verified plans. Each ecosystem uses its selected API model. Break-even is the daily message count where API spend equals the plan price.

Plan Plan $/mo API model Same usage via API Value × Break-even /day Verdict
ChatGPT Go ChatGPT Expanded GPT-5.5 Instant access · may include ads $8.00 GPT-5.6 Sol $26.12 ×3.3 9 plan saves $18.12/mo
ChatGPT Plus ChatGPT GPT-5.6 Sol reasoning · agent mode · Deep Research $20.00 GPT-5.6 Sol $26.12 ×1.3 23 plan saves $6.12/mo
ChatGPT Pro (5×) ChatGPT 5× Plus usage · GPT-5.6 Sol Pro reasoning $100 GPT-5.6 Sol $26.12 ×0.3 110 API saves $73.88/mo
ChatGPT Pro (20×) ChatGPT 20× Plus usage · maximum Codex and Deep Research $200 GPT-5.6 Sol $26.12 ×0.1 230 API saves $174/mo
Claude Pro Claude ~5× Free usage · includes Claude Code $20.00 ($17.00 annual) Claude Opus 4.8 $22.07 ×1.1 27 ≈ break-even
Claude Max 5x Claude 5× Pro usage · higher Claude Code throughput $100 Claude Opus 4.8 $22.07 ×0.2 140 API saves $77.93/mo
Claude Max 20x Claude 20× Pro usage · monthly billing only $200 Claude Opus 4.8 $22.07 ×0.1 270 API saves $178/mo
Google AI Plus Gemini 2× Gemini limits · 400 GB storage $7.99 Gemini 2.5 Pro $8.47 ×1.1 28 ≈ break-even
Google AI Pro Gemini Expanded Gemini 3.1 Pro limits · 5 TB storage $19.99 Gemini 2.5 Pro $8.47 ×0.4 71 API saves $11.52/mo
Google AI Ultra (entry) Gemini Developer-oriented Ultra tier added at I/O 2026 $99.99 Gemini 2.5 Pro $8.47 ×0.1 350 API saves $91.52/mo
Google AI Ultra Gemini Up to 20× Pro limits · cut from $249.99 at I/O 2026 $200 Gemini 2.5 Pro $8.47 ×0.0 710 API saves $192/mo

Value × = what the same volume would bill at API list prices, as a multiple of the plan price. Plans cap usage in rolling windows vendors don't publish in tokens — a very high multiple means you'd likely meet those caps.

How it works

The math behind the verdict

The API side runs your profile through the same engine as our LLM API calculator: list input/output rates, prompt caching with cache-write premiums where vendors charge them, reasoning tokens billed at the output rate. The plan side is a verified price sheet — every plan carries the date it was checked on the vendor's page.

/* API side — per message */
input_rate   = hit × cached_price + (1 − hit) × write_or_base_price
msg_cost     = in_tokens × input_rate / 1M
             + (out_tokens + reasoning_tokens) × output_rate / 1M

/* verdict — per plan */
api_monthly  = msg_cost × msgs_per_day × 30
value        = api_monthly ÷ plan_price      /* >1 → plan is better value */
break_even   = plan_price ÷ (msg_cost × 30)  /* messages per day */

Ecosystem model defaults follow what the consumer app actually serves: ChatGPT plans run on GPT-5.6 Sol, Claude plans on Claude Opus 4.8. Google's app runs Gemini 3.1 Pro, whose API price is still in preview — we price Google usage on Gemini 2.5 Pro, the closest stable rate, which slightly understates the API side. Swap the model in the calculator to test any other equivalence.

Full methodology lives at /methodology/. Found a stale plan price? Report it.

Worked examples

Three verdicts, real numbers

Computed from the current plan sheet and API snapshot — recalculated on every data refresh.

Verdict 01 · casual use
A few questions a day? Keep your $20.
8 short messages a day on GPT-5.6 Sol costs pocket change at API rates — below even ChatGPT Go. Light users subsidize heavy ones on every flat plan.
Usage8 msgs/day · 300 in / 250 out
API cost$2.13/mo
ChatGPT Go$8.00/mo
ChatGPT Plus$20.00/mo
API wins
pay-as-you-go beats every plan at this volume
Verdict 02 · the $100 middle
Power chat lands between the tiers
120 reasoning-heavy messages a day on GPT-5.6 Sol bills $338/mo at API rates — the workload the $100 Pro tier was cut for.
Usage120 msgs/day · +2K thinking
API cost$338/mo
Pro 5× ($100)×3.4 value
Pro 20× ($200)×1.7 value
×3.4
the $100 tier carries this load — $200 buys headroom
Verdict 03 · Claude Code
Full-time agentic coding: Max 20x by a mile
800 agent requests a day at 12K input tokens each — even with 80% cache hits the API bill dwarfs the flat plan.
Usage800 req/day · 12K in · 80% cache
API cost$1,795/mo
Max 20x$200/mo
Break-even89 req/day
×9.0
the plan bills a fraction of the metered equivalent
Beyond the token math

What each side actually buys

01
Subscriptions cap usage in rolling windows

Plus, Pro, and Max tiers are "unlimited subject to guardrails" in marketing copy but throttle in practice — 5-hour windows, weekly ceilings, per-model quotas. Vendors publish multipliers (Max 20x = 20× Pro), never absolute token numbers. A high value multiple here means the plan is a bargain and that you're the user those caps were written for.

02
The API meters every token — and never says no

Pay-as-you-go has no usage wall: heavy days just cost more. You also get programmatic access, model pinning, and batch/caching discounts. What you lose is the app: no Projects, no Deep Research button, no image generation UI — you're building your own harness or running a tool like Claude Code against a key.

03
Subscriptions bundle features the API can't price

Deep Research runs, agent modes, image generation, voice, file uploads, memory — a $20 plan bundles workflows that would each bill separately via API. If you use two or more of them weekly, the plan's effective value is higher than the token math alone.

04
Claude Code runs on either — that's the real fork

The same CLI works on a Pro/Max subscription or an API key. Subscription = flat bill, session caps; API = uncapped, metered per token with prompt caching. The coding presets above model exactly this decision — for full-time agentic work the flat plan usually wins by a wide multiple.

05
Annual billing changes one answer

Claude Pro drops to $17/mo billed annually ($200 up front) — a 15% cut that moves its break-even. As of 2026-07-16, Max and ChatGPT's paid tiers publish no annual discount on the web plans.

Worth it — answered.

The questions this calculator exists for, answered with the current numbers. Every figure recomputes when a vendor reprices.

Q · 01 Is ChatGPT Plus worth it? +
Run your real volume through the calculator — the answer is a number, not an opinion. At typical chat sizes (600 tokens in, 400 out, a 500-token thinking budget), Plus ($20/mo) breaks even around 23 messages a day against GPT-5.6 Sol API rates. Below that, pay-as-you-go is cheaper; above it, Plus wins on price — and it bundles agent mode, Deep Research, and image generation the API doesn't.
Q · 02 Is ChatGPT Pro worth it — and which tier? +
Pro now comes in two prices with identical features: $100/mo for 5× Plus usage, $200/mo for 20×. At the default profile the $100 tier breaks even near 110 messages/day. The honest sequence: outgrow Plus's caps first, take the $100 tier, and move to $200 only when the 5× ceiling appears in your week. Full tier comparison: ChatGPT Plus vs Pro.
Q · 03 Is Claude Pro worth it? +
At $20/mo ($17/mo billed annually) Claude Pro breaks even around 27 messages a day against Claude Opus 4.8 API list prices at the default profile — and it includes Claude Code. Light users are cheaper on the API; anyone chatting daily or coding with it clears the bar comfortably. Tier details: Claude Free vs Pro.
Q · 04 Claude Max 5x or 20x — or just use the API? +
This is really a Claude Code question. The heavy-coding preset (800 requests/day at 12K input tokens each, 80% cache hits) would bill about $1,795/mo at API rates — ×9.0 the price of Max 20x ($200/mo). Max 5x ($100/mo) fits evening-scale agentic work. If you only code occasionally, Pro or a metered API key is the cheaper start. See Claude Pro vs Max.
Q · 05 Is the OpenAI API cheaper than ChatGPT Plus? +
For most people, yes — until it isn't. API billing is metered per token, so a handful of chats a day costs single-digit dollars a month, well under $20. The crossover at typical sizes is roughly 23 messages/day (the calculator recomputes it for your exact profile). Heavy reasoning use flips the answer fast: thinking tokens bill at the output rate and pile up invisibly.
Q · 06 What do Google's AI plans map to? +
Google re-cut its lineup at I/O 2026: AI Plus ($7.99/mo, 2× Gemini limits), AI Pro ($19.99/mo), and two Ultra tiers ($99.99 and $199.99 — the top tier came down from $249.99). The consumer app runs Gemini 3.1 Pro; its API price is still marked preview, so this calculator prices Google usage on Gemini 2.5 Pro rates — the closest stable equivalent — and notes the gap in the methodology.
Q · 07 How is the API cost computed? +
Your messages/day × tokens per message run through the same engine as our LLM API calculator: vendor list prices with prompt caching modeled (cache misses billed at cache-write rates where vendors charge them — e.g. Anthropic's 1.25× write premium) and reasoning tokens billed as output. A flat 30-day month converts daily volume to monthly. Not modeled: enterprise discounts, taxes, retries.
Q · 08 Do subscriptions really have usage limits? +
Yes — every consumer plan throttles: rolling windows (commonly 5-hour), weekly ceilings on the top models, and per-feature quotas. Vendors publish only relative multipliers (Max 20x = 20× Pro; ChatGPT Pro = up to 20× Plus), never absolute token budgets, which is why this page compares the price of your volume rather than pretending to know the caps. If your value multiple runs above ~10×, expect to meet the ceiling.
Deep dive

Deciding between tiers?

Feature-by-feature tier comparisons — limits, model access, and who each tier is actually for.

Embed

Put this calculator on your site

Free to embed — your readers always see current plan prices. The widget links back here for the full plan table.

<iframe
  src="https://aicost.tools/calculator/subscription-vs-api/embed/"
  width="100%" height="820" loading="lazy"
  style="border:1px solid #3d3a32"
  title="Subscription vs API Calculator by AI//COST">
</iframe>
More calculators

The cost toolbox

Every tool runs on the same live-pricing backbone. See the full index →

One weekly digest. Zero noise.

One weekly digest · No spam, ever · Unsubscribe in one click