Every Current OpenAI API Price in One Table
GPT-6 Astra, OpenAI's flagship since early September 2026, costs $10.00 per million input tokens and $50.00 per million output tokens. The cheapest current tier, gpt-5.6-luna, costs $0.20 / $1.20.
Every figure here was checked on 16 September 2026 against the vendors' own published pricing pages.
| Model | Input / 1M | Output / 1M | 1M in + 1M out | Output multiple |
|---|---|---|---|---|
| gpt-6-astra | $10.00 | $50.00 | $60.00 | 5.0x |
| gpt-6-astra above 200k context | $20.00 | $75.00 | $95.00 | 3.75x |
| gpt-5.6-sol | $4.00 | $20.00 | $24.00 | 5.0x |
| gpt-5.6-terra | $2.00 | $12.00 | $14.00 | 6.0x |
| gpt-5.6-luna | $0.20 | $1.20 | $1.40 | 6.0x |
| gpt-5.5 | $5.00 | $30.00 | $35.00 | 6.0x |
| gpt-5.4 | $2.50 | $15.00 | $17.50 | 6.0x |
| gpt-5.4-mini | $0.75 | $4.50 | $5.25 | 6.0x |
| gpt-5.4-nano | $0.20 | $1.25 | $1.45 | 6.25x |
The blended and multiple columns are ours, derived by arithmetic from the two published prices. The spread from top to bottom of the current line is 42.9x on a balanced 1M in / 1M out workload.
One naming note that causes real confusion: the Sol, Terra and Luna suffixes replaced the older Instant, Thinking, mini and nano scheme in July 2026. Anything still labelled gpt-5.6-mini is an outdated reference.
AI Perks tracks $7.7M in credits across 194 companies, which is the other half of this page.

What GPT-6 Pricing Means If You Are On GPT-5
GPT-6 Astra costs 8x the input price and 5x the output price of the legacy gpt-5 it succeeds. On a balanced workload the blended increase is 5.3x.
That is the single number most 2026 budgets have wrong. The jump is not incremental.
Astra is also the first OpenAI model in this line with a published long context surcharge. Above 200k tokens the price roughly doubles on input and rises 50% on output, to $20.00 / $75.00. If your prompts carry large retrieved context, model that tier separately rather than averaging it in.
Three things OpenAI has not published, and which we will not guess at: Astra's maximum context window, any benchmark result, and any rate limit or free tier ceiling. Pages quoting those numbers this month are inventing them.
The Legacy GPT-5 Line Shuts Down 10 December 2026
The gpt-5, gpt-5-mini and gpt-5-nano API endpoints retire on 10 December 2026. Do not start new work on them.
| Legacy model | Input / 1M | Output / 1M | 1M in + 1M out |
|---|---|---|---|
| gpt-5 | $1.25 | $10.00 | $11.25 |
| gpt-5-mini | $0.25 | $2.00 | $2.25 |
| gpt-5-nano | $0.05 | $0.40 | $0.45 |
Here is the under-discussed part. For anyone running gpt-5-nano, the shutdown is a price increase, not a migration. The nearest current equivalent, gpt-5.6-luna at $0.20 / $1.20, is 4x the input price and 3x the output price of nano, which works out to roughly 3.1x blended.
Teams on full gpt-5 get the opposite result. Moving down to gpt-5.6-terra at $2.00 / $12.00 raises input cost but is close on output, and moving to gpt-5.6-luna is dramatically cheaper than what they pay today. Classify your traffic before you pick a landing spot. Credits that absorb a migration quarter are listed at AI Perks.

OpenAI vs Anthropic API Pricing, September 2026
At the top of both lineups the price is identical: GPT-6 Astra and Claude Fable 5.1 both cost $10 / $50 per million tokens. The difference shows up in the middle tiers.
| Model | Input / 1M | Output / 1M | 1M in + 1M out |
|---|---|---|---|
| gpt-6-astra | $10.00 | $50.00 | $60.00 |
| Claude Fable 5.1 | $10.00 | $50.00 | $60.00 |
| Claude Mythos 5.1 | $10.00 | $50.00 | $60.00 |
| gpt-5.5 | $5.00 | $30.00 | $35.00 |
| Claude Opus 5 | $5.00 | $25.00 | $30.00 |
| gpt-5.6-terra | $2.00 | $12.00 | $14.00 |
| Claude Sonnet 5 | $2.00 | $10.00 | $12.00 |
| Claude Haiku 4.5 | $1.00 | $5.00 | $6.00 |
| gpt-5.6-luna | $0.20 | $1.20 | $1.40 |
| Meta Muse Spark 1.1 to 1.3 | $1.25 | $4.25 | $5.50 |
Two patterns fall straight out of the table. Anthropic prices output more cheaply than OpenAI at every matched input price below the top tier, by $5 per million at the $5 input tier and $2 per million at the $2 input tier. At the $10 input tier the two are identical. And OpenAI owns the bottom: luna at $1.40 blended is 4.3x cheaper than Haiku 4.5.
Meta's Muse Spark sits between them at $1.25 / $4.25. There is also a Contributor tier at $0.10 / $0.20, the cheapest number on this page by a wide margin, in exchange for Meta training on your prompts and completions. Meta runs no startup credit program for Muse, so that price is the price. Muse Image is billed per image at $0.01.
We have deliberately omitted Google Gemini. We did not verify Gemini pricing in this research, and a wrong number on a pricing page is worse than a missing one. The same applies to xAI, covered in the FAQ below.
The Claude Sonnet 5 Price Increase That Was Cancelled
Anthropic had scheduled Claude Sonnet 5 to rise from $2 / $10 to $3 / $15 on 1 September 2026, then cancelled it. The $2 / $10 price, originally introductory through 31 August, is now the standard price.
Anthropic's own wording: "The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur."
This is badly under-reported, and it is worth real money. At 100M input and 20M output tokens a month, the cancelled increase means an estimated $400 a month rather than $600, a third off. If you re-platformed off Sonnet 5 earlier this year on the assumption the increase was coming, the assumption no longer holds.
Worth noting alongside it: Claude Opus 5, launched 24 July 2026, shipped at the same $5 / $25 as Opus 4.8. Two consecutive Anthropic releases have now held price flat, against OpenAI's 5.3x jump from gpt-5 to Astra.

What Each Tier Costs Per Month at Real Volume
At 100M input and 20M output tokens per month, GPT-6 Astra costs an estimated $2,000 and gpt-5.6-luna costs $44.
These are estimates derived from the published per-token prices above, with no discounts, caching or batch rates applied.
| Model | Estimated monthly cost |
|---|---|
| gpt-6-astra | $2,000 |
| Claude Fable 5.1 | $2,000 |
| gpt-5.5 | $1,100 |
| Claude Opus 5 | $1,000 |
| gpt-5.6-sol | $800 |
| gpt-5.4 | $550 |
| gpt-5.6-terra | $440 |
| Claude Sonnet 5 | $400 |
| Meta Muse Spark | $210 |
| gpt-5.4-mini | $165 |
| gpt-5.6-luna | $44 |
The gap between the top and bottom rows is $1,956 a month, or $23,472 a year, for the same token volume.
A year of flagship usage at this volume is an estimated $24,000, which a low five figure credit grant covers outright. AI Perks tracks which providers currently grant them.
Which Tier Fits Which Workload
Route by output volume, not by prestige. Output costs 5x to 6.25x input across the current OpenAI line, so the model that writes the most tokens dominates your bill.
High volume classification, extraction and routing. gpt-5.6-luna at $0.20 / $1.20, or gpt-5.4-nano at $0.20 / $1.25. Functionally identical pricing.
Everyday application traffic. gpt-5.6-terra at $2.00 / $12.00 or Claude Sonnet 5 at $2.00 / $10.00. Same input price, Sonnet is $2 cheaper per million out.
Long form generation and agents. gpt-5.6-sol at $4.00 / $20.00 is the honest middle. Sol is $36 cheaper per blended million than Astra.
Genuinely hard work. gpt-6-astra or Claude Fable 5.1, both $10 / $50, and watch the 200k context threshold on Astra.
A two tier router that sends 80% of traffic to luna and 20% to Astra costs an estimated $13.12 per blended million, against $60.00 for flagship-only. That is a 78% reduction from routing alone, before any credits. Stack credits from AI Perks on top and the first year can cost close to nothing.

Frequently Asked Questions
How much does the GPT-6 API cost?
GPT-6 Astra costs $10.00 per million input tokens and $50.00 per million output tokens, verified 16 September 2026. Above a 200k token context the price rises to $20.00 / $75.00. A balanced 1M in plus 1M out request costs $60.00. Credits covering that bill are tracked at AI Perks.
Is GPT-6 more expensive than GPT-5?
Yes, substantially. GPT-6 Astra is 8x the input price and 5x the output price of legacy gpt-5, which is roughly 5.3x on a balanced workload. At 100M in and 20M out per month that is an estimated $2,000 against $325. Budget for the jump before you migrate.
When does the legacy GPT-5 API shut down?
The gpt-5, gpt-5-mini and gpt-5-nano API endpoints retire on 10 December 2026. Do not start new builds on them. Teams on gpt-5-nano should note the nearest current tier costs roughly 3.1x more, so this is a price increase for them, not a like-for-like move.
Did Anthropic raise Claude Sonnet 5 pricing?
No. The increase to $3 / $15 scheduled for 1 September 2026 was cancelled, and $2 / $10 is now the standard price rather than an introductory one. Anthropic states the increase "will not occur." Claude Opus 5 also launched at the same $5 / $25 as Opus 4.8.
What is the cheapest OpenAI model in 2026?
gpt-5.6-luna at $0.20 per million input and $1.20 per million output, which works out to $1.40 for a balanced 1M in plus 1M out request. gpt-5.4-nano is effectively tied at $0.20 / $1.25. On that derived basis luna is 42.9x cheaper than Astra, nano 41.4x. See AI Perks for credits.
How much does the xAI Grok API cost?
We will not print a number. Published sources conflict badly on current xAI token pricing, and pricing varies by model within the Grok line. Check the xAI console for your account's current rates rather than trusting a third party table, including ours. Everything else on this page is vendor-verified.
Know the price, then get someone else to pay it.