GPT-6 COST & PRICING
API pricing, ChatGPT plans and global
subscription prices.
01 / BUILD · API COST
Model
Choose a GPT-6 model. Compare published token prices.
Processing
Select the service tier that matches your needs.
Flex and Batch bill at 50% off Standard rates — same tokens, half the price, lower priority. Fast mode costs 2× Standard where supported. GPT-6 Astra and GPT-6.1 Sol also offer Ultrafast mode at 6× Standard — the fastest tier. Astra Ultrafast is US/global processing only; 6.1 Sol Ultrafast also supports EU data residency. Processing availability is model-specific: EU regional processing supports Fast on Luna, Sol and 6.1 Sol, but not on Astra.
Using Standard processing
Workload presets
Or try a common use case to get started quickly.
These are example workloads, not universal usage assumptions — every number stays editable.
Example workload — adjust these values to match your usage.
Your workload
Enter four values. Your estimate updates automatically.
What you send to the model.
Use the token count from your API logs, or start with an example workload.
What the model generates.
How many API calls you expect each month.
Repeated prompt or context reused across requests.
Input buckets
Fresh input 50% · Cached read 50% · Cache write 0%
Input above a model's long-context threshold uses long rates for the full request. Rates exclude tools, taxes and negotiated discounts.
Cost Summary
Pricing context · Short context
Estimated monthly cost
Cost breakdown
Compare models (same workload)
| Model | Cost | VS SELECTED |
|---|
Per day (30-day month): —
Compare current and previous models on this workload
| Model | Per request | Per month | Per year |
|---|
Ways to lower this estimate
Cost only — model quality is not evaluated. Scenarios are compared separately.
02 / USE
Find a plan based on how you actually use ChatGPT.
USAGE PRESSURE —
Choose your uses and daily active time. Balanced and a 5-day week are selected to start.
BEST FIT TO REVIEW
Why this fit
Consider the next tier only if:
SUBSCRIPTION REFERENCE
Complete Plan Fit to see a suggested plan reference.
This compares raw cost equivalents, not subscription entitlements.
Subscriptions include app features, tools, files, images and eligible Work / Codex access, with their own usage policies. API token charges do not measure those entitlements.
FEATURE SNAPSHOT NEEDS REVIEW · Verify official access before deciding.
Based on published plan features and an estimated usage profile. Exact limits vary by account, feature and usage pattern.
Daily active time × 5 or 7 days produces an approximate weekly band, not an allowance. For an open-ended 8+ hour band, we use its lower bound. Limits and reset schedules differ by feature; a typical week is only a way to summarize your activity.
Our editorial rules use weekly midpoint bands below 2.5, 5, 10, 20 and 40 hours, then 40 or more. To prevent abrupt jumps, the weekly time signal is capped at half a step above the selected daily band. Tool-dependent image use alone is capped at HIGH pressure and a Pro 100 review candidate; coding, research and Work / Codex can add more. We use the highest workload plus one small, capped mixed-use adjustment, never multiply selected uses. Faster means a preference for performance and headroom, with a small editorial bias. Pro 500 review requires all three editorial gates: VERY HIGH pressure, Faster preference and our highest overall high-demand threshold. Sustained use is an additional signal: coding or research needs a weekly lower bound of at least 40 active hours; Work / Codex may use a weekly midpoint of 40 or more. That midpoint alone never qualifies a profile. These signals never measure plan capacity. Balanced remains at Pro 200, with Pro 500 guidance only for real limits, maximum included usage or Ultrafast. Save allowance favors lower suitable tiers and does not enter this Pro 500 path. Everyday chat alone never triggers a paid-plan recommendation. These are editorial thresholds, not OpenAI quota measurements.
Verified specific features override editorial estimates: Astra Ultrafast requires at least Pro 500 for personal use. Team + Ultrafast requires eligible Enterprise / Edu review. Documented speed-mode allowance behavior applies only to supported modes and scopes in Work / Codex, not to choosing Faster in general Chat. No message, token, task or hour quota is inferred, and time is never converted into API tokens. Real account behavior should outweigh the estimate: Never does not trigger an upgrade, Sometimes suggests headroom, and Often suggests reviewing the next suitable tier without promising to remove limits. At Pro 500, review usage patterns, credits for eligible features or eligible workspace options instead of an unavailable higher personal tier.
Manual feature snapshot; re-check official sources before changing plans.
03 / BUY · APPLE APP STORE
ChatGPT subscription pricing · Separate from API billing
Apple App Store · Monitored storefront list prices
Loading the latest saved observations…
Observed storefront list prices. USD estimates use dated reference FX rates.
Taxes, eligibility and checkout totals may differ. Data sources / methodology →
Short-context Standard rates per 1M tokens. Input above 272K uses long-context rates for the full request. Fast costs 2×; Flex and Batch cost 0.5× where supported.
| Model | Input / 1M | Cached read / 1M | Cache write / 1M | Output / 1M |
|---|---|---|---|---|
| GPT-6 Luna | $0.10 | $0.01 | $0.125 | $0.50 |
| GPT-6.1 Sol | $2.00 | $0.10 | $2.50 | $10.00 |
| GPT-6 Astra | $10.00 | $1.00 | $12.50 | $50.00 |
| Model | Input / 1M | Cached read / 1M | Cache write / 1M | Output / 1M |
|---|---|---|---|---|
| GPT-6 Sol · Previous | $2.00 | $0.20 | $2.50 | $10.00 |
Prices last verified: October 10, 2026 · Source: OpenAI platform pricing (verify at developers.openai.com/api/docs/pricing)
OpenAI pricing · verified October 10, 2026 · methodology →
Output tokens cost more than input.
Repeated prompt prefixes use discounted cached-input pricing.
Standard / Fast / Flex / Batch change effective price where supported.
OpenAI bills GPT-6 by the token, split into input and output. Input is everything you send — system prompt, conversation history, the user message, tool definitions. Output is everything the model generates, including hidden reasoning tokens. Output is priced higher because generation is the expensive part.
A RAG chatbot with long retrieved context might run 10:1 input-to-output. A code generator might run 1:3. Two apps with the same request count can differ 10× in cost. Before optimizing the model choice, measure your real token split — most developers overestimate output and underestimate how much history they resend every turn.
If your system prompt and context repeat across requests, cached input pricing (a fraction of standard input rates) applies to the repeated prefix. Apps with stable system prompts can cut input spend substantially with caching — the exact figure depends on the cached share. In the calculator above, move the cached-input slider to see the effect on your workload.
Processing modes are configured per model. Fast costs 2× Standard and combines with EU data residency on Luna, Sol and 6.1 Sol (Astra excluded). Standard is full price at full speed. Flex costs half but your requests queue behind Standard traffic — fine for background jobs, bad for user-facing latency. Batch also runs 50% off with up to 24-hour turnaround for non-urgent bulk work. Ultrafast (Astra and 6.1 Sol) costs 6× Standard for the lowest latency OpenAI sells — with EU data residency available on 6.1 Sol only. Toggle the tier in the calculator to see the difference on your workload.
Luna is the lowest-priced tier ($0.10 input / $0.50 output per 1M): classification, extraction, simple chat, bulk processing. GPT-6.1 Sol costs 20× Luna for fresh input and output, with a 5% cached-read rate — whether that premium is justified depends on your quality evals, not on this page. Astra is the highest-priced GPT-6 tier ($10.00 input / $50.00 output per 1M), 5× 6.1 Sol for fresh input and output; this page cannot say whether the extra cost buys better output.
Quick answers to the pricing questions developers actually search for.
Compare current models in the calculator above using your complete workload. Relative costs depend on caching, context band, mode and region.
GPT-6 vs Claude cost comparison · GPT-6 vs Gemini cost comparison. Provider cache mechanics and processing availability differ; unverified cache-write rates are unavailable in the calculator.
GPT-6.1 Sol vs GPT-6 Sol compares current and previous Sol without assuming a universal 50% saving.
Dedicated calculators with worked examples for each GPT-6 tier — plus head-to-head cost comparisons.