Grok Imagine Image 2.0: auto means low, then medium

With auto quality, Grok Imagine Image 2.0 bills as low when you generate and as medium when you edit, so the same "auto" setting costs more on an edit. The prices below come from a third-party pricing breakdown, not from Sume.

The price list

Per image, by quality and size, plus a flat fee for each input image you send.

• Each input image adds $0.01.
• auto = low for generation, medium for editing.

What that means for an edit

An edit sends at least one input image, so it pays the medium rate plus $0.01 per input. At 1K that is $0.07 for one input image against $0.04 for a plain low-quality generation. If you only wanted a quick variation, set quality explicitly rather than leaving it on auto.

How to check the price on Sume

Sume publishes each image model's capabilities and billable lines in its catalog. List models with GET /v1/images/models, then read one model's pricing array from GET /v1/images/models/{model_id}/endpoints. I did not find a Grok image model named in the Sume image docs, so treat availability and price as something the live catalog answers, not this post.

curl "https://api.sume.com/v1/images/models" \
  -H "Authorization: Bearer $SUME_API_KEY"

# then, for a model id you found in the list:
curl "https://api.sume.com/v1/images/models/{model_id}/endpoints" \
  -H "Authorization: Bearer $SUME_API_KEY"

Auto is a different thing on Sume

On Sume, model: "sume/auto" lets Sume choose the family and never discloses which one ran. Separately, ChatGPT Image 2.5's quality: auto reserves max according to the Sume image docs, the opposite instinct to xAI's low-for-generation default. Whenever a vendor says auto, read what it resolves to before you budget.