OpenAI API pricing
As of November 14, 2026, the OpenAI API costs $0.05 to $10 per million input tokens and $0.4 to $50 per million output tokens, across 23 ChatGPT models. GPT-6 Luna costs $0.1 input and $0.5 output; GPT-4o mini costs $0.15 input and $0.6 output; GPT-5.4 nano costs $0.2 input and $1.25 output; GPT-5.4 mini costs $0.75 input and $4.5 output; GPT-6.1 Terra costs $1 input and $5 output; o3 costs $2 input and $8 output; GPT-6.1 Sol costs $2 input and $10 output; GPT-4o costs $2.5 input and $10 output. A typical request of 1,000 input and 500 output tokens costs $0.00025 on GPT-5 nano and $0.035 on GPT-6 Astra Law.
OpenAI API price per 1M tokens, every model
| Model | Input | Cached input | Cache write | Output | Batch input | Batch output | Context | Typical request |
|---|---|---|---|---|---|---|---|---|
GPT-5 nanogpt-5-nano | $0.05 | $0.005 | – | $0.4 | $0.025 | $0.2 | 272K | $0.00025 |
GPT-6 Lunagpt-6-luna | $0.1 | $0.01 | $0.125 | $0.5 | $0.05 | $0.25 | 922K | $0.00035 |
GPT-4o minigpt-4o-mini | $0.15 | $0.075 | – | $0.6 | $0.075 | $0.3 | 128K | $0.00045 |
GPT-5.6 Lunagpt-5.6-luna | $0.2 | $0.02 | $0.25 | $1.2 | $0.1 | $0.6 | 922K | $0.0008 |
GPT-5.4 nanogpt-5.4-nano | $0.2 | $0.02 | – | $1.25 | $0.1 | $0.625 | 272K | $0.000825 |
GPT-4.1 minigpt-4.1-mini | $0.4 | $0.1 | – | $1.6 | $0.2 | $0.8 | 1.05M | $0.0012 |
GPT-5 minigpt-5-mini | $0.25 | $0.025 | – | $2 | $0.125 | $1 | 272K | $0.00125 |
GPT-5.4 minigpt-5.4-mini | $0.75 | $0.075 | – | $4.5 | $0.375 | $2.25 | 272K | $0.003 |
GPT-6.1 Terragpt-6.1-terra | $1 | $0.05 | $1.25 | $5 | $0.5 | $2.5 | 922K | $0.0035 |
GPT-4.1gpt-4.1 | $2 | $0.5 | – | $8 | $1 | $4 | 1.05M | $0.006 |
o3o3 | $2 | $0.5 | – | $8 | $1 | $4 | 200K | $0.006 |
GPT-5gpt-5 | $1.25 | $0.125 | – | $10 | $0.625 | $5 | 272K | $0.00625 |
GPT-5.1gpt-5.1 | $1.25 | $0.125 | – | $10 | $0.625 | $5 | 272K | $0.00625 |
GPT-6 Solgpt-6-sol | $2 | $0.2 | $2.5 | $10 | $1 | $5 | 922K | $0.007 |
GPT-6.1 Solgpt-6.1-sol | $2 | $0.1 | $2.5 | $10 | $1 | $5 | 922K | $0.007 |
GPT-4ogpt-4o | $2.5 | $1.25 | – | $10 | $1.25 | $5 | 128K | $0.0075 |
GPT-5.6 Terragpt-5.6-terra | $2 | $0.2 | $2.5 | $12 | $1 | $6 | 922K | $0.008 |
GPT-5.2gpt-5.2 | $1.75 | $0.175 | – | $14 | $0.875 | $7 | 272K | $0.00875 |
GPT-5.4gpt-5.4 | $2.5 | $0.25 | – | $15 | $1.25 | $7.5 | 1.05M | $0.01 |
GPT-5.6gpt-5.6 | $4 | $0.4 | $5 | $20 | $2 | $10 | 922K | $0.014 |
GPT-5.5gpt-5.5 | $5 | $0.5 | – | $30 | $2.5 | $15 | 1.05M | $0.02 |
GPT-6 Astragpt-6-astra | $10 | $1 | $12.5 | $50 | $5 | $25 | 922K | $0.035 |
GPT-6 Astra Lawgpt-6-astra-law | $10 | $1 | – | $50 | – | – | 922K | $0.035 |
US dollars per million tokens, sorted by the cost of a typical request of 1,000 input and 500 output tokens. Cache write is the standard (5-minute) cache. Also listed as gpt-5.6-sol (same as GPT-5.6).
How much does an OpenAI API request cost?
A typical request of 1,000 input and 500 output tokens costs $0.00035 on GPT-6 Luna, $0.00045 on GPT-4o mini, $0.000825 on GPT-5.4 nano, $0.003 on GPT-5.4 mini, $0.0035 on GPT-6.1 Terra, $0.006 on o3, $0.007 on GPT-6.1 Sol and $0.0075 on GPT-4o. A coding agent step with 50K tokens of context and 2K output tokens costs $0.15 on GPT-4o.
| Model | Short chat message 300 in, 200 out | Typical request 1K in, 500 out | Long document summary 20K in, 1K out | Coding agent step 50K in, 2K out | 1,000 typical requests |
|---|---|---|---|---|---|
| GPT-6 Luna | $0.00013 | $0.00035 | $0.0025 | $0.006 | $0.35 |
| GPT-4o mini | $0.000165 | $0.00045 | $0.0036 | $0.0087 | $0.45 |
| GPT-5.4 nano | $0.00031 | $0.000825 | $0.00525 | $0.0125 | $0.83 |
| GPT-5.4 mini | $0.00112 | $0.003 | $0.0195 | $0.0465 | $3.00 |
| GPT-6.1 Terra | $0.0013 | $0.0035 | $0.025 | $0.06 | $3.50 |
| o3 | $0.0022 | $0.006 | $0.048 | $0.12 | $6.00 |
| GPT-6.1 Sol | $0.0026 | $0.007 | $0.05 | $0.12 | $7.00 |
| GPT-4o | $0.00275 | $0.0075 | $0.06 | $0.15 | $7.50 |
At standard rates, for the newest model of each ChatGPT family. A token is about 4 characters of English text, so 1,000 tokens is roughly 750 words.
How batch, prompt caching and long context change the price
Batch API. 22 of the 23 priced ChatGPT models have a batch API price. Requests sent as a batch are processed within 24 hours and cost 50% less: GPT-4o drops from $2.5 input and $10 output to $1.25 and $5 per million tokens. See batch API pricing.
Prompt caching. All 23 priced ChatGPT models charge less for cached input, the part of a prompt that repeats across requests: typically 90% less. GPT-4o charges $1.25 per million cached input tokens instead of $2.5. Writing a prompt to the cache costs extra: 1.25× the input price for the standard cache (GPT-6 Astra: $12.5 per million tokens). The write pays for itself on the first request that reads the cache. See prompt caching pricing.
Long context. ChatGPT models accept 128K to 1.05M input tokens. Long prompts cost more on 10 models: GPT-6 Luna charges $0.2 input and $0.75 output per million tokens once a prompt is over 272K tokens, instead of $0.1 input and $0.5 output; GPT-5.6 Luna charges $0.4 input and $1.8 output per million tokens once a prompt is over 272K tokens, instead of $0.2 input and $1.2 output; GPT-6.1 Terra charges $2 input and $7.5 output per million tokens once a prompt is over 272K tokens, instead of $1 input and $5 output; GPT-6 Sol charges $4 input and $15 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $10 output; GPT-6.1 Sol charges $4 input and $15 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $10 output; GPT-5.6 Terra charges $4 input and $18 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $12 output; GPT-5.4 charges $5 input and $22.5 output per million tokens once a prompt is over 272K tokens, instead of $2.5 input and $15 output; GPT-5.6 charges $8 input and $30 output per million tokens once a prompt is over 272K tokens, instead of $4 input and $20 output; GPT-5.5 charges $10 input and $45 output per million tokens once a prompt is over 272K tokens, instead of $5 input and $30 output; GPT-6 Astra charges $20 input and $75 output per million tokens once a prompt is over 272K tokens, instead of $10 input and $50 output. A 372K-token prompt with 2K output tokens on GPT-6 Astra costs $7.59, against $3.82 at the standard rate. The other 13 models list one price for their whole context window.
OpenAI API price changes
Since October 22, 2026, when apicost.app started tracking, no OpenAI model has changed its standard API price. 2 new models were added, most recently GPT-6 Astra Law on Nov 1, 2026. 9 models left the price list, most recently GPT-3.5 Turbo on Oct 24, 2026.
- Nov 1, 2026: GPT-6 Astra Law added to the price list at $10 input and $50 output per 1M tokens.
- Oct 27, 2026: GPT-6.1 Terra added to the price list at $1 input and $5 output per 1M tokens.
- Oct 24, 2026: GPT-3.5 Turbo left the price list (it cost $0.5 input and $1.5 output per 1M tokens).
- Oct 24, 2026: GPT-3.5 Turbo 16k left the price list (it cost $3 input and $4 output per 1M tokens).
- Oct 24, 2026: GPT-4 left the price list (it cost $30 input and $60 output per 1M tokens).
- Oct 24, 2026: GPT-4 1106 Preview left the price list (it cost $10 input and $30 output per 1M tokens).
Dated from the day a change reached the price list. Every provider: LLM API price changes.
OpenAI API vs Claude and Gemini
It depends on the model. The cheapest OpenAI model, GPT-5 nano, costs $0.00025 per typical request; the cheapest Claude model, Claude Haiku 5.5, costs $0.00035. GPT-4o ($0.0075) is closest in price to Claude Sonnet 5 ($0.007): about the same.
| ChatGPT model | Input / output | Typical request | Closest Claude model | Closest Gemini model |
|---|---|---|---|---|
| GPT-6 Luna | $0.1 / $0.5 | $0.00035 | Claude Haiku 5.5 $0.1 / $0.5 · $0.00035 · about the same | Gemini 2.5 Flash Lite $0.1 / $0.4 · $0.0003 · 1.2× cheaper |
| GPT-4o mini | $0.15 / $0.6 | $0.00045 | Claude Haiku 5.5 $0.1 / $0.5 · $0.00035 · 1.3× cheaper | Gemini 2.5 Flash Lite $0.1 / $0.4 · $0.0003 · 1.5× cheaper |
| GPT-5.4 nano | $0.2 / $1.25 | $0.000825 | Claude Haiku 5.5 $0.1 / $0.5 · $0.00035 · 2.4× cheaper | Gemini 3.1 Flash Lite $0.25 / $1.5 · $0.001 · 1.2× the price |
| GPT-5.4 mini | $0.75 / $4.5 | $0.003 | Claude Haiku 4.5 $1 / $5 · $0.0035 · 1.2× the price | Gemini 3.6 Flash $0.75 / $3.75 · $0.00263 · 1.1× cheaper |
| GPT-6.1 Terra | $1 / $5 | $0.0035 | Claude Haiku 4.5 $1 / $5 · $0.0035 · about the same | Gemini 3.6 Flash $0.75 / $3.75 · $0.00263 · 1.3× cheaper |
| o3 | $2 / $8 | $0.006 | Claude Sonnet 5 $2 / $10 · $0.007 · 1.2× the price | Gemini 3.5 Flash $1.5 / $9 · $0.006 · about the same |
| GPT-6.1 Sol | $2 / $10 | $0.007 | Claude Sonnet 5 $2 / $10 · $0.007 · about the same | Gemini 4 Argon $2 / $10 · $0.007 · about the same |
| GPT-4o | $2.5 / $10 | $0.0075 | Claude Sonnet 5 $2 / $10 · $0.007 · about the same | Gemini 3.1 Pro Preview $2 / $12 · $0.008 · about the same |
For each ChatGPT model, the model from the other provider whose typical request costs the nearest, with its input / output price per million tokens, its typical request, and how that compares with the ChatGPT model.
Questions and answers
How much does the OpenAI API cost?
As of November 14, 2026, the OpenAI API costs $0.05 to $10 per million input tokens and $0.4 to $50 per million output tokens, across 23 ChatGPT models. GPT-6 Luna costs $0.1 input and $0.5 output; GPT-4o mini costs $0.15 input and $0.6 output; GPT-5.4 nano costs $0.2 input and $1.25 output; GPT-5.4 mini costs $0.75 input and $4.5 output; GPT-6.1 Terra costs $1 input and $5 output; o3 costs $2 input and $8 output; GPT-6.1 Sol costs $2 input and $10 output; GPT-4o costs $2.5 input and $10 output. A typical request of 1,000 input and 500 output tokens costs $0.00025 on GPT-5 nano and $0.035 on GPT-6 Astra Law.
Did OpenAI raise its API prices?
No. Since October 22, 2026, when apicost.app started tracking, no OpenAI model has changed its standard API price. 2 new models were added, most recently GPT-6 Astra Law on Nov 1, 2026. 9 models left the price list, most recently GPT-3.5 Turbo on Oct 24, 2026.
What is the cheapest OpenAI model?
GPT-5 nano (gpt-5-nano), at $0.05 per million input tokens and $0.4 per million output tokens.
Is the OpenAI API free?
You pay per token: the OpenAI API is pay as you go, billed separately from any ChatGPT subscription. The cheapest model, GPT-5 nano, costs $0.05 per million input tokens and $0.4 per million output tokens, so 1,000 typical requests cost $0.25.
Is the OpenAI API cheaper than a ChatGPT subscription?
It depends on how much you use it. $20 of OpenAI API usage pays for about 2,857 typical requests (1,000 input and 500 output tokens) on GPT-6.1 Sol, or 80,000 on GPT-5 nano. At 30 typical requests a day, a month costs $6.30 on GPT-6.1 Sol and $0.22 on GPT-5 nano. Long chats cost more per message than that, because every turn sends the whole conversation again as input tokens. The API is billed separately from any ChatGPT app subscription.
Why is the OpenAI API so expensive?
Mostly because of output tokens and long prompts. GPT-6.1 Sol charges 5× more for output ($10 per million tokens) than for input ($2). Reasoning models bill their thinking as output tokens, even though you don't see it. And a chat sends the whole conversation again as input on every turn, so a long chat gets more expensive with each message. To pay less: GPT-5 nano is 28× cheaper per typical request than GPT-6.1 Sol, the batch API takes 50% off and cached input costs 90% less.
How much does an OpenAI API request cost?
A typical request of 1,000 input and 500 output tokens costs $0.00035 on GPT-6 Luna, $0.00045 on GPT-4o mini, $0.000825 on GPT-5.4 nano, $0.003 on GPT-5.4 mini, $0.0035 on GPT-6.1 Terra, $0.006 on o3, $0.007 on GPT-6.1 Sol and $0.0075 on GPT-4o. A coding agent step with 50K tokens of context and 2K output tokens costs $0.15 on GPT-4o.
Is the OpenAI batch API cheaper?
Yes. 22 of the 23 priced ChatGPT models have a batch API price. Requests sent as a batch are processed within 24 hours and cost 50% less: GPT-4o drops from $2.5 input and $10 output to $1.25 and $5 per million tokens.
How much does prompt caching save on the OpenAI API?
All 23 priced ChatGPT models charge less for cached input, the part of a prompt that repeats across requests: typically 90% less. GPT-4o charges $1.25 per million cached input tokens instead of $2.5. Writing a prompt to the cache costs extra: 1.25× the input price for the standard cache (GPT-6 Astra: $12.5 per million tokens). The write pays for itself on the first request that reads the cache.
Does long context cost more on the OpenAI API?
Yes, on some models. ChatGPT models accept 128K to 1.05M input tokens. Long prompts cost more on 10 models: GPT-6 Luna charges $0.2 input and $0.75 output per million tokens once a prompt is over 272K tokens, instead of $0.1 input and $0.5 output; GPT-5.6 Luna charges $0.4 input and $1.8 output per million tokens once a prompt is over 272K tokens, instead of $0.2 input and $1.2 output; GPT-6.1 Terra charges $2 input and $7.5 output per million tokens once a prompt is over 272K tokens, instead of $1 input and $5 output; GPT-6 Sol charges $4 input and $15 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $10 output; GPT-6.1 Sol charges $4 input and $15 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $10 output; GPT-5.6 Terra charges $4 input and $18 output per million tokens once a prompt is over 272K tokens, instead of $2 input and $12 output; GPT-5.4 charges $5 input and $22.5 output per million tokens once a prompt is over 272K tokens, instead of $2.5 input and $15 output; GPT-5.6 charges $8 input and $30 output per million tokens once a prompt is over 272K tokens, instead of $4 input and $20 output; GPT-5.5 charges $10 input and $45 output per million tokens once a prompt is over 272K tokens, instead of $5 input and $30 output; GPT-6 Astra charges $20 input and $75 output per million tokens once a prompt is over 272K tokens, instead of $10 input and $50 output. A 372K-token prompt with 2K output tokens on GPT-6 Astra costs $7.59, against $3.82 at the standard rate. The other 13 models list one price for their whole context window.
Is the OpenAI API cheaper than Claude?
It depends on the model. The cheapest OpenAI model, GPT-5 nano, costs $0.00025 per typical request; the cheapest Claude model, Claude Haiku 5.5, costs $0.00035. GPT-4o ($0.0075) is closest in price to Claude Sonnet 5 ($0.007): about the same.
Source: OpenAI pricing, via the LiteLLM price list.