Skip to main content

Cost estimator for usage-based pricing

Estimate the cost of using GPT for Work in usage-based pricing.

info

If you use the Agent, you need to add the cost of the Agent messages to the cost of the executions below.

Choose your model

OpenAI gpt-4o
No API key needed
gpt-4o
$30 / 1M tokens
gpt-4o-mini
$2 / 1M tokens
gpt-5-search-api
$40 / 1M tokens + $10 / 1K executions
gpt-5.1
$40 / 1M tokens
gpt-5
$40 / 1M tokens
gpt-5-mini
$8 / 1M tokens
gpt-5-nano
$1 / 1M tokens
gpt-5-chat
$20 / 1M tokens
gpt-4.1
$20 / 1M tokens
gpt-4.1-mini
$6 / 1M tokens
gpt-4.1-nano
$1 / 1M tokens
o1
$70 / 1M tokens
o3
$10 / 1M tokens
o4-mini
$10 / 1M tokens
gpt-3.5-turbo
$2.5 / 1M tokens
sonar
$10 / 1M tokens + $5 to $12 / 1K executions
sonar-reasoning
$10 / 1M tokens + $5 to $12 / 1K executions
gemini-2.5-flash
$4 / 1M tokens + $35 / 1K executions
Anthropic
claude-4-sonnet
$20 / 1M tokens
Anthropic
claude-4.5-sonnet
$20 / 1M tokens
Anthropic
claude-4.5-haiku
$6 / 1M tokens
Use your API key 🔑
gpt-4o 🔑
$1 / 1M tokens
gpt-4o-mini 🔑
$0.3 / 1M tokens
gpt-4o-search-preview 🔑
$1 / 1M tokens
gpt-4o-mini-search-preview 🔑
$0.3 / 1M tokens
gpt-5-search-api 🔑
$1 / 1M tokens
gpt-5.1 🔑
$1 / 1M tokens
gpt-5 🔑
$1 / 1M tokens
gpt-5-mini 🔑
$1 / 1M tokens
gpt-5-nano 🔑
$1 / 1M tokens
gpt-5-chat 🔑
$1 / 1M tokens
gpt-4.1 🔑
$1 / 1M tokens
gpt-4.1-mini 🔑
$1 / 1M tokens
gpt-4.1-nano 🔑
$1 / 1M tokens
gpt-4-turbo 🔑
$1 / 1M tokens
gpt-4 🔑
$1 / 1M tokens
o1 🔑
$1 / 1M tokens
o3-mini 🔑
$1 / 1M tokens
o3 🔑
$1 / 1M tokens
o4-mini 🔑
$1 / 1M tokens
gpt-3.5-turbo 🔑
$1 / 1M tokens
ft:gpt-3.5-turbo 🔑
$1 / 1M tokens
ft:gpt-4o 🔑
$1 / 1M tokens
ft:gpt-4o-mini 🔑
$0.3 / 1M tokens
sonar 🔑
$1 / 1M tokens
sonar-pro 🔑
$1 / 1M tokens
sonar-reasoning 🔑
$1 / 1M tokens
sonar-reasoning-pro 🔑
$1 / 1M tokens
grok-3 🔑
$1 / 1M tokens
grok-3-mini 🔑
$1 / 1M tokens
grok-3-fast 🔑
$1 / 1M tokens
grok-3-mini-fast 🔑
$1 / 1M tokens
gemini-2.5-flash 🔑
$1 / 1M tokens
gemini-2.5-flash-lite 🔑
$1 / 1M tokens
gemini-2.5-pro 🔑
$1 / 1M tokens
gemini-2.0-flash 🔑
$1 / 1M tokens
gemini-2.0-flash-lite 🔑
$1 / 1M tokens
Anthropic
claude-4-sonnet 🔑
$1 / 1M tokens
Anthropic
claude-4.5-sonnet 🔑
$1 / 1M tokens
Anthropic
claude-4.5-haiku 🔑
$1 / 1M tokens
deepseek-reasoner 🔑
$1 / 1M tokens
mistral-large 🔑
$1 / 1M tokens
mistral-small 🔑
$1 / 1M tokens
ministral-8b 🔑
$1 / 1M tokens
ministral-3b 🔑
$1 / 1M tokens
openrouter 🔑
$1 / 1M tokens
$30 / 1M tokens

Tokens & executions → $$$

Executions needed

One execution = one cell of output

Average tokens per execution

Average number of tokens consumed by a single execution (including both your prompt and the model’s response). You can calculate tokens for a typical execution in the estimator below.

Tokens needed

200000

Balance needed

$6.00
Buy a pack

Tokens per execution estimator

0 characters0 words0 tokens
Your input should include your full prompt and custom instructions if they exist.
0 characters0 words0 tokens
This is the output returned by the AI. Note that for reasoning models, the amount of output tokens can actually be significantly higher than your visible output because of reasoning tokens.

Tokens per execution

0
You can estimate token counts for various providers with our tokenizer tool.

Detailed execution estimator

Function

0 characters0 words0 tokens$0

=GPT(prompt, [value])

0 tokens$0
0 characters0 words0 tokens$0
0 characters0 words0 tokens$0
This is the response returned by the AI. Note that it will usually be slightly different from what you see in your spreadsheet as our functions do some post-processing to improve formatting.
$0

Notes

  • Without an API key: The estimated cost covers everything and is deducted from your balance. Models that support prompt caching will get a 75% discount on cached input tokens, so the estimated cost can be higher than your actual cost.
  • With an API key: You pay GPT for Work the estimated cost (deducted from your balance) and also pay the AI provider directly for API usage.
For more information about the models, see AI providers and models. Note that token counts can vary from one AI provider to another. For some models (such as Sonar, Mistral, and Deepseek) the token count is approximate. (The OpenAI tokenizer is used)

Unlimited BYOK usage subscription

Annual plan with custom pricing. Pay per user, not per token, with unlimited usage. Requires your own API key or endpoint.
Contact sales