Ideogram V3 Turbo, Balanced, Quality: how Sume maps them

On Sume you do not pass Turbo, Balanced or Quality directly. You send quality as low, medium or high on ideogram/ideogram-v3, and Sume converts it to the provider's TURBO, BALANCED or QUALITY rendering speed. If you leave quality out, the request runs as BALANCED.

This is read from the Image API docs and the request-mapping code, 2026-09-29. Nothing here describes how the tiers look or how long they take, because Sume's docs make no such claim.

Which quality word gives which speed?

The mapping is one fixed table in the image request code. The provider words are also the ones the SDK's quality hint lists for this model: ideogram: TURBO|BALANCED|QUALITY.

Can I send TURBO or QUALITY directly?

The provider adapter accepts the provider words too, in any letter case, and treats turbo like low, balanced like medium and quality like high. An unrecognised value falls back to BALANCED rather than failing. The public contract is still low, medium and high, so use those in code you plan to keep.

Does the tier change the price?

No. The price table holds one rate for ideogram-v3: $0.06 list, shown as $0.075 on the endpoint pricing line after Sume's 1.25 margin. The reserve rounds up to whole cents, so one image reserves $0.08 at any tier. The reasoning behind those numbers is in Ideogram 4.0 price per image.

How do I run the same prompt at all three tiers?

Send three requests that differ only in quality, and keep the same aspect_ratio. This model lists fifteen ratios, including 4:5.

for q in low medium high; do
  curl -s -X POST https://api.sume.com/v1/images \
    -H "Authorization: Bearer $SUME_API_KEY" \
    -H "Content-Type: application/json" \
    -d "{
      \"model\": \"ideogram/ideogram-v3\",
      \"prompt\": \"A poster reading OPEN LATE in bold letters\",
      \"aspect_ratio\": \"4:5\",
      \"quality\": \"$q\"
    }"
done

Which tier should I start with?

Because the three tiers cost the same, the choice is about your own workflow rather than your wallet. A sensible pattern is to leave quality out while you tune the prompt, since that already runs BALANCED, then send the final wording once at high. Compare the outputs yourself: Sume's docs describe the mapping, not a visual or timing difference, so this post does not rank the tiers. Each generation completes or fails as a whole, and a failed one is not billed, so a comparison run is cheap to abandon.

What happens if I pass a parameter Ideogram V3 does not list?

Sume returns 400 unsupported_parameter. This model has no resolution, for example, so quality is the only tier knob. Check GET /v1/images/models to see which parameters a model lists, and read Ideogram 4.0 API for the rest of the request shape.