---
format: "aidr-story-markdown/v1"
id: "c63748f2bb70e84b11e7033d0f00fad191e490d8d8ff539ce32112d4afbaab7a"
canonical_url: "https://aidr.today/c63748f2?lang=en"
title: "OpenAI Ships Images 2.5 and API Models That Take #1 and #2 in 3 Image Arenas"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-09-08T20:52:22.000Z"
category: "Releases"
topics: ["openai","images-25","gpt-image-25","multimodal","image-generation","benchmark"]
source_urls: ["https://huggingnews.com/ai/openai-ships-images-25-and-api-models-that-take-1-and-2-in-3-image-arena-7cd102e1","https://x.com/OpenAI/status/2097394962782384318","https://x.com/OpenAI/status/2097394956457623964","https://x.com/OpenAIDevs/status/2097399267866648769","https://x.com/OpenAIDevs/status/2097399305808335188","https://x.com/arena/status/2097400515546255754","https://x.com/sama/status/2097410967978324010"]
summary: "All ChatGPT, ChatGPT Work, and Codex users now have access to the Images 2.5 update across desktop, mobile, and web platforms. The release enables comment-based edits to modify specific elements of an image and enhances fidelity for textures and lighting. OpenAI also launched GPT-Image-2.5 Flare and Sunburst via its API, with the Flare model cutting generation latency by 50% compared to GPT-Image-2. Sunburst and Flare occupy the #1 and #2 spots respectively in the Text-to-Image, Image Edit, and Multi-Image Edit Arenas. Sunburst's performance improved by 81 points in the Multi-Image Edit Arena and 40 points in Text-to-Image relative to the GPT-Image-2 medium model. CEO Sam Altman stated the tools are highly capable but cannot solve difficult math problems."
---

# OpenAI Ships Images 2\.5 and API Models That Take \#1 and \#2 in 3 Image Arenas

> [Open the canonical story](<https://aidr.today/c63748f2?lang=en>)

**Published:** 2026-09-08T20:52:22.000Z
**Category:** Releases
**Topics:** openai, images\-25, gpt\-image\-25, multimodal, image\-generation, benchmark

## Summary

All ChatGPT, ChatGPT Work, and Codex users now have access to the Images 2\.5 update across desktop, mobile, and web platforms\. The release enables comment\-based edits to modify specific elements of an image and enhances fidelity for textures and lighting\. OpenAI also launched GPT\-Image\-2\.5 Flare and Sunburst via its API, with the Flare model cutting generation latency by 50% compared to GPT\-Image\-2\. Sunburst and Flare occupy the \#1 and \#2 spots respectively in the Text\-to\-Image, Image Edit, and Multi\-Image Edit Arenas\. Sunburst's performance improved by 81 points in the Multi\-Image Edit Arena and 40 points in Text\-to\-Image relative to the GPT\-Image\-2 medium model\. CEO Sam Altman stated the tools are highly capable but cannot solve difficult math problems\.

## Sources

- [Story source](<https://huggingnews.com/ai/openai-ships-images-25-and-api-models-that-take-1-and-2-in-3-image-arena-7cd102e1>)
- [Story source](<https://x.com/OpenAI/status/2097394962782384318>)
- [Supporting source](<https://x.com/OpenAI/status/2097394956457623964>)
- [Supporting source](<https://x.com/OpenAIDevs/status/2097399267866648769>)
- [Supporting source](<https://x.com/OpenAIDevs/status/2097399305808335188>)
- [Supporting source](<https://x.com/arena/status/2097400515546255754>)
- [Supporting source](<https://x.com/sama/status/2097410967978324010>)