GPT-4 and GPT-4o in Sparkforge Workflows

By Sparkforge Support Team ·Last reviewed May 3, 2026

Sparkforge workflows include a GPT action that runs GPT-3.5, GPT-4, GPT-4o, or GPT-4o Mini against your prompt. Pricing is pass-through token-based at OpenAI's rates with no Sparkforge markup.

The GPT powered by OpenAI workflow action lets a workflow generate text on the fly — follow-up messages, conversation summaries, email drafts, social captions — and feed the output into the next step. You pick the model per step, write a prompt, and pipe contact merge fields into it.

What's available

Model switcher on every AI step: GPT-3.5, GPT-4, GPT-4 Turbo, GPT-4o, and GPT-4o Mini.
Optional system prompt for tone and behavior guidance.
Merge field support for personalization ({{contact.first_name}}, custom fields, event notes).

Pricing

The GPT action bills at zero markup. You pay OpenAI's exact token rates — input tokens (your prompt) plus output tokens (the response) — against the model you pick. The old flat per-execution fee is gone for these models.

Per-million-token rates

GPT-4o: $2.50 input / $10.00 output
GPT-4o Mini: $0.15 input / $0.60 output

For rough scale: $10 of GPT-4o handles roughly 3 million input words or 750,000 output words. The same $10 on GPT-4o Mini handles roughly 50 million input words or 12.5 million output words.

Picking a model

Model	When to use
GPT-3.5 Turbo	Cheap, fast, good for short rewrites where accuracy isn't critical.
GPT-4 Turbo	Complex prompts where you need consistent reasoning.
GPT-4o	Balanced speed and quality. Good default.
GPT-4o Mini	Highest volume, lowest cost. Routine messages and summaries.

Set up a GPT step

In your workflow, add the GPT powered by OpenAI action.
Pick an Action Type: Analyze Sentiment, Summarize, Translate, or Custom.
If Custom, write your prompt. Drop merge fields in with the custom value picker, e.g. Write a friendly follow-up to {{contact.first_name}} about {{contact.custom_value.service}}.
Save and test the workflow against a real contact.

Where the output goes

The generated text is available to any later step. Common patterns:

Pipe it into an Email or SMS step as the body.
Write it to a custom field on the contact.
Send it to a Slack channel or webhook for review before it goes out.

Example prompts

Use case	Prompt
Lead follow-up	Draft a friendly message to {{contact.name}} thanking them for their interest in {{contact.custom_value.service}}.
Conversation summary	Summarize this conversation in 2-3 bullet points.
Promo email	Write a persuasive email promoting our {{contact.custom_value.offer}} to a potential client.

Frequently Asked Questions

Which Sparkforge plans include the GPT workflow action?

Workflow automation lives in Standard ($465/mo) and Pro ($730/mo). Token usage on top is billed pass-through at OpenAI's rates.

Can I mix models in the same workflow?

Yes. Each AI step picks its own model independently. Run a cheap GPT-4o Mini summarization step and a higher-quality GPT-4 Turbo drafting step in the same flow.

How many words is a million tokens?

Roughly 750,000 words, give or take. OpenAI's tokenizer averages about 0.75 words per token, but real numbers vary by language and formatting.

Where do I see what GPT usage is costing me?

Token usage and cost roll up in your AI Billing settings. Costs scale with input length, output length, and the model you choose — there's no per-execution fee on the new models.