GPT Image 2.5 for packaging text: default high, xhigh and max prices

For dense packaging text on ChatGPT Image 2.5 through Sume, set quality and image_size explicitly. If you omit quality, Sume uses high. At 1024x1024 the Sume docs give provider output costs of $0.09366 for xhigh and $0.21072 for max, before input tokens and Sume pricing. Times 1.25 that is about $0.117 and $0.263 per image.

The token rates behind it

The docs state that both GPT Image 2.5 ids, openai/gpt-image-2.5 and openai/gpt-image-2.5-sunburst, use the same Fal token rates: $30 per million output image tokens, $8 per million input image tokens and $5 per million input text tokens. Input token counts are estimates and Fal rounds the total up to $0.0001.

What you get for the extra

Sume's Image 1.0 docs say to use a higher quality for finals, dense text or packaging, and the same logic applies here. It is guidance, not a measured result, so run your own label copy at medium and xhigh and compare. On 12 test labels, xhigh at about $0.117 costs about $1.40 and medium costs less.

Size matters as much as quality

Leaving image_size out makes Sume reserve the upper bound of output tokens. A text-to-image call with no size or quality quotes $0.2225 on Sume, so a 1024x1024 xhigh call at $0.117 is the cheaper request even though it asks for more quality. Custom sizes need both edges as multiples of 16, a maximum edge of 3840, an aspect ratio of at most 3:1 and 655,360 to 8,294,400 pixels.

Alternatives for the same job

Nano Banana 2.1 is $0.10 at 1K and Google says it is built for accurate text rendering. Imagen 4 Ultra is $0.075 but cannot edit. Pick by testing, using the 12-prompt label test.

A quick cost grid for 100 labels

Using the output-only estimate, 100 labels at xhigh 1024x1024 come to about 100 x $0.117075 = $11.71, plus input tokens. At max they come to about 100 x $0.2634 = $26.34. The default no-size, no-quality request quotes $0.2225 each, so 100 of those are $22.25.

That ordering is the reason to set both fields. The cheapest sensible request for dense text is often the explicit xhigh at a modest size, not the default. Check by asking Sume for a quote on your exact settings before you queue the batch.

The two ids, Flare (openai/gpt-image-2.5) and Sunburst (openai/gpt-image-2.5-sunburst), use the same token rates per the docs. Auto routing uses Flare.

Ask for a quote on a single call with your exact settings, then multiply. The golden numbers in this post come from the Sume pricing code and the docs. Your prompt adds a few text input tokens, which are priced at $5 per million, so a 200-token prompt adds about $0.001 at the provider. That is small next to the image cost, but reference images add input image tokens at $8 per million, and those can matter in an edit.

Always set output_format explicitly if your pipeline expects PNG, and read the usage.cost field in the response, which the docs define as the USD amount that Sume bills to your wallet.