# GPT-Image 2.5 in Luma Agents

By Steven Van · 2026-09-09

Luma Agents gain GPT-Image 2.5 with two modes, Flare for speed and Sunburst for precise reference-based edits that carry into video.

[Luma AI](<https://creatorstoolbox.com/tools/luma-ai>) has added GPT-Image 2.5 to its Agents, available in two modes. Flare is built for speed and volume, while Sunburst is aimed at precise edits and maximum fidelity: bring a reference image, change what's off, keep the rest, then carry the result into video.

The [announcement](<https://x.com/i/status/2097763347743601096>) frames Sunburst's edit-and-keep workflow as a bridge into Luma's video tools, letting a reference image be refined before it becomes a video generation.

[View media](<https://video.twimg.com/amplify_video/2097762860487028744/vid/avc1/1280x720/kDKFKdEXYovpPfah.mp4>)

![Luma AI](<https://is1-ssl.mzstatic.com/image/thumb/Purple211/v4/a6/0b/27/a60b279d-5298-626c-d35e-6179d483b216/AppIcon-1x_U007ephone-0-1-0-85-220-0.png/512x512bb.jpg>)

Luma AI

Scan real objects and scenes with a phone camera into photorealistic 3D models using NeRF, or generate 3D assets and video with Dream Machine.

[View Luma AI →](<https://creatorstoolbox.com/tools/luma-ai>)

Sources

[Original source](<https://x.com/i/status/2097763347743601096>)

[Read GPT-Image 2.5 in Luma Agents on Creators Toolbox](<https://creatorstoolbox.com/blog/luma-ai-gpt-image-2-5-in-luma-agents>)

---
Canonical source: https://creatorstoolbox.com/blog/luma-ai-gpt-image-2-5-in-luma-agents