MetaSkill/skills/le-gpt-image/references/gallery/openai-image-2.5-migration.md

GPT Image 2.5: migration and evaluation

Read only for model migration or an authorized comparison, not routine image requests. Sources checked 2026-09-09; locally organized official-source notes, not a new MIT Cookbook snapshot or locally verified image results.

What changes from Image 2

• OpenAI positions Flare around speed and Sunburst around demanding quality. Both generate and edit; Sunburst is not edit-only.
• If Image 2 already passes your quality bar, evaluate Flare first. Otherwise, evaluate Sunburst. Compare the same inputs and explicit settings before rewriting prompts; keep the quality setting fixed in the first comparison. Establish acceptable output quality before optimizing latency. Official migration guidance
• xhigh and max are options, not automatic upgrades. Higher quality may not improve a particular task. Test against acceptance criteria and budget. Official quality guidance

Quoted text, layout constraints and reference invariants are shared techniques already covered in craft.md, not newly invented 2.5 prompt syntax. Keep parameter details in models.md, not in task prompts.

Local evaluation checklist

Only run comparisons or extra variants when authorized. Before a trial, write down what would make the output unusable; use the selected task slice's checks. Record model, full prompt, attached inputs, dimensions, quality, format, elapsed time, reported usage, output path and pass/fail reason. Keep failed attempts in the cost accounting. A documentation example is not our measured result.

Source boundaries

• Current official prompting guide: illustrated 2.5 workflows. Its complete runnable example still selects Image 2; do not copy that model or batch count over the user's confirmed choices.
• Historical local Cookbook: captured 2026-04-23, with its existing source/license retained. The official notebook checked on 2026-09-09 still targets Image 2. Its recommendations and executable defaults are not current 2.5 settings.
• Community templates: attributed adaptations with author-reported source models, not OpenAI recommendations or verified outputs.