How to Use GPT-5 in Sparkforge Workflows

By Sparkforge Support Team ·Last reviewed May 3, 2026

Add an OpenAI GPT action to any Sparkforge workflow and pick a GPT-5 variant. Pricing matches OpenAI's published per-token rates and bills through your Sparkforge AI wallet.

Sparkforge workflows can call OpenAI's GPT-5 family directly. Pick a model, write a prompt, map fields from earlier steps, and the action returns generated text the rest of the workflow can use.

Model options

Model	Best for	Context window	Input / 1M tokens	Output / 1M tokens
GPT-5	Long-form reasoning, document analysis, high-stakes copy	256k	$1.25	$10.00
GPT-5 Mini (default)	Email drafting, chat replies, marketing copy	128k	$0.25	$2.00
GPT-5 Nano	High-volume classification, short replies, light summarization	64k	$0.05	$0.40
GPT-5.1	Stronger instruction following, conversational tone	128k	$1.25	$10.00

Pricing mirrors OpenAI's published rates and is billed through your Sparkforge AI wallet. Agency markup and client rebilling apply automatically.

Token math

Sparkforge counts input tokens (prompt and context) and output tokens (the response). One token is roughly four characters of English text. Charges prorate to the nearest thousandth of a million tokens.

Example: a 400-word email is about 2,000 characters or 500 tokens. Sent through GPT-5 Mini with a 500-token reply:

Input: 500 × $0.25 / 1M = $0.000125
Output: 500 × $2.00 / 1M = $0.001
Total: $0.001125 per action

Add a GPT-5 step

Open the workflow and click the + icon where you want the AI step.
Search for GPT powered by OpenAI and select the GPT action.
In the Model dropdown, keep the GPT-5 Mini default or pick GPT-5, GPT-5 Nano, or GPT-5.1.
Compose the prompt and map any inbound field tokens.
Save the step and publish the workflow.

Prompt tips

Open the system prompt with the model's role and the exact output format you want.
Keep instructions specific. "Reply in two sentences" beats "keep it short."
For classification or extraction, supply two or three labeled examples in the prompt.
Pin temperature low for deterministic tasks; raise it for creative copy.

Frequently Asked Questions

Which GPT-5 model is the default?

GPT-5 Mini is the default. It balances speed, quality, and cost for most workflow tasks.

How am I billed for GPT-5 usage?

Sparkforge bills per token at OpenAI's published rates through your AI wallet. Agency markup and client rebilling apply automatically.

What is a token?

A token is roughly four characters of English text. A 400-word email is about 500 tokens. Both your prompt and the response count toward the total.

Which model handles long documents?

GPT-5 supports a 256k context window, the largest of the four. Use it for long-form reasoning, document analysis, and high-stakes copy.

Can I switch models without rebuilding the workflow?

Yes. Change the Model dropdown on the GPT step and republish. The prompt and field mappings stay intact.