OpenAI API pricing (2026): what each GPT model costs per token, and how to pay less
updated ยท prices checked on vendor pages
quick answer: OpenAI API pricing, as of nov 1, 2026, per million tokens (input / output): GPT-6.1 Sol is $2 / $10, GPT-6.1 Terra is $1 / $5, GPT-6 Luna is $0.10 / $0.50 and GPT-6 Astra is $10 / $50, for prompts up to 272K tokens. longer prompts cost more. batch and flex halve those prices, fast mode doubles them, and cached input is a fraction of the normal rate.
the OpenAI API is #3 of 10 on our LLM APIs & models board (8.5/10). it's the biggest ecosystem going, and the GPT-6 range runs from cheap-as-chips Luna to big-brain Astra. it's not our #1 for coding agents (that's the Claude API), but it's the safe default for general-purpose and multimodal apps. here's what it costs.
the price per model
as of nov 1, 2026, from OpenAI's API pricing page, per million tokens, short context (up to 272K input tokens):
| model | input | output | cached input | batch / flex (input / output) |
|---|---|---|---|---|
| GPT-6 Astra | $10 | $50 | $1 | $5 / $25 |
| GPT-6.1 Sol | $2 | $10 | $0.10 | $1 / $5 |
| GPT-6.1 Terra | $1 | $5 | $0.05 | $0.50 / $2.50 |
| GPT-6 Luna | $0.10 | $0.50 | $0.01 | $0.05 / $0.25 |
| GPT-6 Sol | $2 | $10 | $0.20 | $1 / $5 |
long context (over 272K input tokens) costs roughly double on input: GPT-6.1 Sol is $4 / $15, GPT-6.1 Terra $2 / $7.50, GPT-6 Luna $0.20 / $0.75 and GPT-6 Astra $20 / $75.
older models are still listed, often at a higher price than the new ones: GPT-5.6 Sol is $4 / $20 (OpenAI says its promotional pricing runs at least until nov 21, 2026), GPT-5.5 is $5 / $30 and GPT-5.4 is $2.50 / $15. so GPT-6.1 Sol is cheaper per token than the GPT-5 generation it replaces. if you're still on an old model id, check the table before you assume it's the bargain.
which model should you pay for?
- the default for most apps and agents: GPT-6.1 Sol at $2 / $10. same list price as GPT-6 Sol, half the cached input price.
- high-volume work that still needs to think a bit: GPT-6.1 Terra at $1 / $5. half of Sol.
- classifying, extracting, cheap sub-agents: GPT-6 Luna at $0.10 / $0.50. twenty times cheaper than Sol on input.
- the hard problems: GPT-6 Astra at $10 / $50. five times Sol, so route the hard steps there, not the whole loop.
a rough sum, our arithmetic on the list prices: an agent run that reads 1 million input tokens and writes 100,000 output tokens costs about $3 on GPT-6.1 Sol ($2 + $1), $1.50 on GPT-6.1 Terra, $0.15 on GPT-6 Luna and $15 on GPT-6 Astra, with no caching and every request under 272K tokens.
how to pay less
- cached input. repeated prompt prefixes are billed at the cached rate: $0.10 instead of $2 on GPT-6.1 Sol, $0.01 instead of $0.10 on GPT-6 Luna. the GPT-6 models also list a price for writing to the 30-minute and 12-hour cache (on GPT-6.1 Sol: $2.50 and $5 per million tokens). for an agent that resends the same system prompt and repo context every turn, this is the big one.
- batch. half price on input and output for jobs that can wait.
- flex. the same half price as batch on the GPT-6 models.
- stay under 272K tokens per request. past that, the long context price kicks in for the request.
- the right model per step. Luna for the grunt work, Sol or Terra for the loop, Astra only where it earns it.
the extras that show up on the bill
- fast mode: double the standard price for faster output (GPT-6.1 Sol is $4 / $20). it's what OpenAI used to call priority processing, renamed on jul 30, 2026. there's also an ultrafast tier for GPT-6 Astra ($60 / $300) and GPT-6.1 Sol ($12 / $60).
- data residency: regional processing endpoints add 10% for models released on or after mar 5, 2026. from nov 1, 2026, FedRAMP endpoints add 10% too.
- web search: $10 per 1,000 calls, plus the search content tokens at the model's rate.
- hosted shell and code interpreter: billed per 20-minute container session, from $0.03 (1 GB) to $1.92 (64 GB).
- file search: $0.10 per GB per day of storage (1 GB free), plus $2.50 per 1,000 tool calls.
- built-in tools still cost tokens: they're billed at the chosen model's per-token rates. the responses, chat completions, realtime, batch and assistants APIs aren't priced separately.
- embeddings: text-embedding-3-small is $0.02 per million tokens, text-embedding-3-large $0.13. moderation (omni-moderation-latest) is free.
- fine-tuning is on the way out: OpenAI says the fine-tuning platform no longer takes new users.
is the OpenAI API free?
OpenAI's pricing page doesn't list a free tier, so we don't claim one: plan on paying per token from the first request. the one model it lists as free is the moderation model.
if you want OpenAI's coding agent in your terminal on a monthly plan rather than per token, that's Codex: see Codex pricing.
OpenAI API vs the other LLM APIs
- the one to beat for coding agents and tool use: Anthropic Claude API, #1 (9.0/10). see Claude API pricing
- the value pick with a free tier: Google Gemini API, #2 (8.5/10). see Gemini API pricing
- GPT, Claude and Gemini on one key: OpenRouter, #4. see OpenRouter pricing
- much cheaper per token: DeepSeek API, #6 (8.0/10)
the rest of the field
the table: the rest of the field
| rank | tool | best for | free tier | paid from | numptr score |
|---|---|---|---|---|---|
| #3 | OpenAI API | general-purpose and multimodal apps | unknown | GPT-6.1 Sol $2/$10 per M tokens | 8.5/10 |
| #1 | Anthropic Claude API | coding agents, tool use | small amount of free credits for new API users | Sonnet 5.5 $2/$10 per M tokens | 9.0/10 |
| #2 | Google Gemini API | long-context and agentic tasks on a budget | free tier on Flash models (Argon not on free tier) | Gemini 4 Argon $2/$10 per M tokens through nov 30, 2026, then $4/$20 | 8.5/10 |
| #4 | OpenRouter | prototyping across providers | 25+ free models, 50 requests a day | pay-as-you-go at model list prices, 5.5% fee on credit purchases | 8.5/10 |
| #6 | DeepSeek API | cheap agent and coding workloads on a tight token budget | unknown: no free tier stated (the pricing page mentions a granted balance) | deepseek-flash $0.30/$1.20 per M tokens at peak, $0.15/$0.60 off-peak | 8.0/10 |
faq
how much does the OpenAI API cost?
per million tokens, as of nov 1, 2026, for prompts up to 272K tokens: GPT-6.1 Sol $2 input / $10 output, GPT-6.1 Terra $1 / $5, GPT-6 Luna $0.10 / $0.50, GPT-6 Astra $10 / $50. batch and flex are half price.
what's the cheapest OpenAI model on the API?
of the GPT-6 range, GPT-6 Luna at $0.10 per million input tokens and $0.50 per million output tokens, or $0.05 / $0.25 on batch. older models like gpt-5-nano are listed at $0.05 / $0.40.
is the OpenAI API free?
the pricing page doesn't list a free tier, so assume you pay per token. the moderation model is the one listed as free.
how much is GPT-6.1 Sol on the API?
$2 per million input tokens and $10 per million output tokens up to 272K input tokens, $4 / $15 above that. cached input is $0.10, batch is $1 / $5 and fast mode is $4 / $20.
is the batch API cheaper on OpenAI?
yes, half price on input and output. flex gives the same price on the GPT-6 models.
tools in this guide
sources
checked 2026-11-01
- OpenAI API pricing (per-model standard, cached input, cache write, long context, batch, flex, fast and ultrafast prices, the 272K short/long context threshold, older model prices, GPT-5.6 Sol promotional pricing date, fast mode rename, data residency and FedRAMP 10% uplift, web search, containers, file search, built-in tool tokens, embeddings, free moderation, fine-tuning wind-down): https://developers.openai.com/api/docs/pricing
- ranks, scores, verdicts and plans: numptr data/tools.json (LLM APIs & models board)