OpenAI GPT Image 2.5: Sunburst and Flare Explained

OpenAI Releases GPT Image 2.5 on 8 September 2026

On 8 September 2026, OpenAI released ChatGPT Images 2.5 across ChatGPT, ChatGPT Work, and Codex, and published two API model variants for developers: GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. The release follows GPT Image 2 and marks OpenAI's most significant image generation update since that model launched. OpenAI reports that image generation latency has dropped by up to 50 per cent compared to Images 2.0 — a meaningful improvement for production workflows where generation time directly affects end-user experience. On the independent Text-to-Image Arena benchmark, which uses Elo scoring from human pairwise preference votes, Sunburst holds first place with 1,421 points, Flare is second with 1,399, the previous GPT Image 2 at medium quality scores 1,381, and Microsoft's MAI-Image-2.6 sits at 1,331.

Flare Versus Sunburst: Two Models for Different Workflows

OpenAI has split GPT Image 2.5 into two distinct API models so developers can choose the appropriate trade-off between speed and precision. Flare is the speed-optimised variant, designed for workflows where generation time is a constraint: real-time product visualisation in consumer apps, batch processing pipelines, and user-facing image flows where latency affects perceived quality. Sunburst is the precision-optimised variant, designed for detailed creative and editing work where consistency across a series matters more than speed. Generation times for Sunburst are longer than Flare, but the model maintains visual coherence across multiple edits and multi-image sequences. OpenAI's stated use cases for Sunburst are campaign creative, product imagery, and image sequences featuring one recurring character — workflows where a faster but less consistent model produces unusable variation between individual outputs.

New ChatGPT Product Features in Images 2.5

Alongside the two API models, GPT Image 2.5 ships four new product features inside ChatGPT and ChatGPT Work. Sketch allows users to draw a rough visual reference directly in the ChatGPT interface, replacing text-only prompting with a spatial layout reference the model uses for composition guidance. Templates provide pre-built starting formats for common creative outputs including posters, merchandise mockups, and product photos, reducing prompt engineering effort for marketing teams using ChatGPT for creative production. Image Comments allow targeted feedback on a specific region of an image rather than requiring a full regeneration from a modified text prompt — directly addressing the most common friction in AI-assisted image editing workflows. Prompt Sharing lets users publish a prompt so others can reuse the same creative brief with their own subjects or details, introducing a community creative layer to the ChatGPT image product.

Where GPT Image 2.5 Sits Competitively

Sunburst's lead over Microsoft's MAI-Image-2.6 by 90 Elo points on the Text-to-Image Arena is substantial in those ranking terms and consistent with its positioning as the highest-precision option for commercial image work currently available via a major API. The 50 per cent latency reduction across both Flare and Sunburst compared to Images 2.0 addresses a competitive gap that previously made GPT Image less suitable for time-sensitive integrations. OpenAI's decision to expose both variants through the same API surface — with model selection as a parameter rather than separate product tiers — gives development teams the flexibility to use different models for different use cases within a single integration, without maintaining separate API credentials or product arrangements.

What GPT Image 2.5 Means for Indian Product and Creative Teams

India's digital economy includes a large base of e-commerce, marketing, and SaaS product teams that use AI image generation for visual output at scale. For engineering teams building products that include image generation — catalogue builders for e-commerce platforms, marketing automation tools, or creative management systems — GPT Image 2.5's dual-model API is directly relevant. The Flare variant suits user-facing product flows where generation speed is part of the user experience. The Sunburst variant suits back-office batch creation of campaign imagery, product renders, and template-based creative where consistency across a series matters. India's e-commerce sector — where visual merchandising at scale is a continuous operational requirement across thousands of SKUs — stands to benefit from both the quality improvements and the latency reduction. Teams that previously routed image generation through slower specialist models now have a more competitive API option that integrates within an existing OpenAI authentication arrangement, reducing integration overhead. For Indian digital marketing agencies and content production teams serving large consumer brands, the Sketch and Image Comments features lower the barrier to iterative AI-assisted creative production for clients who previously found the text-only prompting model too imprecise for professional campaign work.

The Bottom Line

OpenAI released GPT Image 2.5 on 8 September 2026 with two API variants: Flare for speed-sensitive workflows and Sunburst for precision and consistency in detailed creative work. Image generation latency dropped by up to 50 per cent compared to Images 2.0, and Sunburst holds first place on the Text-to-Image Arena at 1,421 Elo — 90 points ahead of its nearest competitor. New ChatGPT product features include Sketch for spatial layout references, Templates for common creative formats, Image Comments for targeted feedback on specific image regions, and Prompt Sharing for community creative reuse. For Indian product teams building e-commerce, marketing automation, or content tools, GPT Image 2.5 offers a more competitive API option for both real-time user-facing and batch back-office image generation within a single OpenAI integration.