# OpenAI API pricing: GPT-6 models, caching, batch, tools and usage tiers

What each GPT model costs on OpenAI's API, the long-context surcharge, cache and batch discounts, the speed tiers, the tools, and what each usage tier allows.

Published 2026-10-28 by Richard Kaminsky and Mitchell Lipyansky, the co-founders of Poly. Canonical: https://usepoly.co/openai-api-pricing
Poly is a multiplayer AI coding workspace: a shared room where your team works with one AI agent, together. Free to start: https://usepoly.co/

**OpenAI API prices per million tokens, input and output: GPT-6 Astra costs $10 and $50, GPT-6.1 Sol $2 and $10, GPT-6.1 Terra $1 and $5, and GPT-6 Luna $0.10 and $0.50. Cached input costs a tenth of the input price or less, the Batch and Flex tiers halve everything, and Fast mode doubles it. Prompts over 272,000 input tokens cost twice as much for input and 1.5 times as much for output. Web search is $10 per 1,000 calls, and new accounts start on a Free usage tier with a $100 monthly limit.** Data residency endpoints add 10% on models released since March 2026.

## Key takeaways

- **Top model:** GPT-6 Astra at $10 and $50, with a 1.05 million token context window.
- **Everyday:** GPT-6.1 Sol at $2 and $10; GPT-6.1 Terra, new on October 26, 2026, at $1 and $5.
- **Cheapest:** GPT-6 Luna at $0.10 and $0.50.
- **Long prompts:** above 272,000 input tokens, input doubles and output rises 1.5 times.
- **Speed:** Batch and Flex 50% off; Fast mode 2x; Ultrafast 6x on Astra and GPT-6.1 Sol.
- **Promotion:** GPT-5.6 Sol at $4 and $20 at least through November 21, 2026.

*OpenAI API prices per million tokens, standard tier, prompts up to 272K tokens, from OpenAI's pricing docs, October 28, 2026.*

| Model | Input | Cached input | Output | Batch (input / output) |
| --- | --- | --- | --- | --- |
| GPT-6 Astra | $10 | $1 | $50 | $5 / $25 |
| GPT-6.1 Sol | $2 | $0.10 | $10 | $1 / $5 |
| GPT-6 Sol | $2 | $0.20 | $10 | $1 / $5 |
| GPT-6.1 Terra | $1 | $0.05 | $5 | $0.50 / $2.50 |
| GPT-6 Luna | $0.10 | $0.01 | $0.50 | $0.05 / $0.25 |
| GPT-5.6 Sol (promotional) | $4 | $0.40 | $20 | $2 / $10 |
| GPT-5.3-Codex | $1.75 | $0.175 | $14 | $0.875 / $7 |

## How much does the OpenAI API cost?

From OpenAI's pricing docs ([OpenAI](https://developers.openai.com/api/docs/pricing)), per million tokens of input, cached input and output:

- **GPT-6 Astra:** $10, $1, $50.
- **GPT-6.1 Sol:** $2, $0.10, $10.
- **GPT-6 Sol:** $2, $0.20, $10. GPT-6.1 Sol costs the same with cheaper cached input.
- **GPT-6.1 Terra:** $1, $0.05, $5. OpenAI calls it "a fast, efficient model for visual reasoning, precise computer use, and clear, natural writing" ([OpenAI](https://developers.openai.com/api/docs/models)).
- **GPT-6 Luna:** $0.10, $0.01, $0.50.

All four GPT-6 models have a 1.05 million token context window and up to 128,000 output tokens.

The GPT-5.6 family is still on sale: GPT-5.6 Sol at $4, $0.40 and $20 on promotional pricing "at least through November 21, 2026", GPT-5.6 Terra at $2, $0.20 and $12, and GPT-5.6 Luna at $0.20, $0.02 and $1.20. Older models cost more than their GPT-6 replacements: GPT-5.5 is $5 and $30, GPT-5.4 $2.50 and $15, and the Pro models $30 and $180. GPT-5.3-Codex, which Codex uses for code review, is $1.75 and $14.

## How much more do long prompts cost?

Requests over 272,000 input tokens are priced higher for the whole request: input and cached input cost twice as much, and output 1.5 times. GPT-6 Astra becomes $20 input, $2 cached and $75 output; GPT-6.1 Sol becomes $4 and $15. Anthropic doesn't charge extra for long context on its current models apart from Haiku 5.5 ([Claude API pricing](/claude-api-pricing)).

## How does caching work on the OpenAI API?

When a request starts with a prefix OpenAI has seen recently, the repeated part is billed at the cached-input price: $1 instead of $10 on Astra, $0.10 instead of $2 on GPT-6.1 Sol. Writing to the cache now has its own price too: 1.25 times input for a 30-minute cache, or 2.5 times for 12 hours. A long system prompt or document reused across many calls costs a fraction of sending it fresh each time, which is also why coding agents like Codex stay affordable over long sessions.

## What do Batch, Flex, Fast and Ultrafast cost?

- **Batch and Flex:** 50% off standard prices. Batch returns results asynchronously; Flex trades speed and availability for price.
- **Fast mode, formerly Priority:** twice the standard price.
- **Ultrafast:** six times the standard price, on GPT-6 Astra and GPT-6.1 Sol only. Astra Ultrafast is $60 input and $300 output.

## What do OpenAI's tools cost?

- **Web search:** $10 per 1,000 calls, plus the tokens of the content it returns.
- **File search:** $2.50 per 1,000 calls; storage $0.10 per GB a day, with 1 GB free.
- **Containers** for code execution: from $0.03 per 20-minute session.
- **Images:** gpt-image-2 and gpt-image-2.5 at $8 per million image input tokens and $30 per million image output tokens, half price in Batch.
- **Voice:** gpt-realtime-2.1 at $32 per million audio input tokens and $64 output; gpt-transcribe at $0.0045 a minute.

Data residency endpoints and FedRAMP endpoints cost 10% more, on models released on or after March 5, 2026. OpenAI is winding down its fine-tuning platform.

## What are the OpenAI API usage tiers?

Each tier sets a monthly usage limit, and moving up depends on what you've paid in total ([OpenAI](https://developers.openai.com/api/docs/guides/rate-limits)):

- **Free:** up to $100 a month, in eligible countries.
- **Build:** after $5 in purchases, up to $500 a month.
- **Launch:** after $100, up to $5,000.
- **Grow:** after $500, up to $200,000.

Rate limits per minute rise with each tier and differ by model.

## API key or a ChatGPT plan?

For Codex, a ChatGPT plan is usually cheaper for daily use: Plus at $20 covers a few focused sessions a week, and Pro at $100 to $500 removes the five-hour limit. With an API key (`codex login --with-api-key`) you pay these token prices and lose cloud tasks, GitHub code review and integrations. Plans and limits are in [Codex pricing](/codex-pricing). For teams, ChatGPT Enterprise's usage-based option bills at this same rate card ([ChatGPT Enterprise pricing](/chatgpt-enterprise-pricing)).

## OpenAI API vs Claude API prices

- **Top models:** GPT-6 Astra at $10 and $50 against Claude Opus 5.5 at $4 and $20, and Fable 5.1 at $10 and $50.
- **Mid-range:** GPT-6.1 Sol and Claude Sonnet 5.5 both at $2 and $10.
- **Small:** GPT-6 Luna and Claude Haiku 5.5 both at $0.10 and $0.50.

Which model does better at coding, at what cost per task, is in [the best AI model for coding](/best-ai-model-for-coding) and [Fable 5.1 vs GPT-6 Astra](/fable-5-1-vs-gpt-6-astra).

## GPT without an API key

Poly (usepoly.co) runs GPT, Claude, Grok and Muse in a shared room where a team works with one AI together, billed as credits per person: no API keys, no usage tiers. GPT-6 Luna is free to start. [What is Poly?](/what-is-poly)

## Common questions

**How much does the OpenAI API cost?**

Per million input and output tokens: GPT-6 Astra $10 and $50, GPT-6.1 Sol $2 and $10, GPT-6.1 Terra $1 and $5, and GPT-6 Luna $0.10 and $0.50. Batch halves the price; Fast mode doubles it.

**How much does GPT-6 Astra cost?**

$10 per million input tokens, $1 cached and $50 output. Above 272,000 input tokens it is $20, $2 and $75. Batch is half price, and Ultrafast six times the standard price.

**What is the OpenAI API free tier?**

New accounts in eligible countries start on the Free usage tier, which has a $100 monthly usage limit. Once an account has paid $5 it moves to Build, with a $500 monthly limit.

**Does OpenAI charge more for long prompts?**

Yes. A request over 272,000 input tokens costs twice as much for input and cached input and 1.5 times as much for output, for the whole request.

**Is the OpenAI API cheaper than ChatGPT Plus for Codex?**

For daily use, usually not: Plus costs $20 a month for a few focused sessions a week. An API key bills every token and drops cloud tasks, GitHub code review and integrations.

More guides: https://usepoly.co/guides · Security: https://usepoly.co/security · Pricing: https://usepoly.co/pricing