OpenAI Launches ChatGPT Images 2.5, Cuts Generation Latency by Up to 50%
OpenAI has released ChatGPT Images 2.5, an upgrade to its image generation system that cuts latency by up to 50% compared to Images 2.0 and improves multi-turn editing accuracy. The company also launched two API-accessible models, GPT-Image-2.5 Flare and Sunburst, for developers.
OpenAI has released ChatGPT Images 2.5, an update to its image generation technology that the company says delivers sharper details, faster generation, and more precise editing than its predecessor, released in April.
The update follows ChatGPT Images 2, which introduced 2K resolution output, multiple aspect ratio support, web-research-informed generation, non-Latin language text rendering, and two intelligence tiers (Instant and Thinking). Images 2.5 builds on that foundation rather than replacing it wholesale.
What's new
According to OpenAI, Images 2.5 produces "more natural lighting and richer textures," is better at preserving subjects across reference photos, and follows multi-turn editing instructions more reliably. The company claims image generation latency has been reduced by up to 50% compared with Images 2.0, allowing faster iteration on concepts.
The release also adds a new "Sketch" feature that lets users draw directly inside ChatGPT and use that drawing as a generation reference. OpenAI is additionally rolling out templates for popular image formats and a prompt-sharing feature that lets users publish image-generation prompts for others to try.
Images 2.5 is available today across ChatGPT, ChatGPT Work, and Codex, on all subscription tiers, across desktop, mobile, and web, per OpenAI.
Two new API models
Developers gain access to two distinct model variants through OpenAI's API:
- GPT-Image-2.5 Flare — positioned as the default option for most applications. OpenAI claims it delivers higher-quality output than GPT-Image-2 at 50% lower latency, targeting use cases like social content, product imagery, visual search, and high-volume generation.
- GPT-Image-2.5 Sunburst — aimed at premium visual workflows requiring tighter edit control, such as production-ready campaign creative or polished product imagery.
OpenAI has not disclosed pricing for either API model at this time. No specific benchmark scores were published alongside the release; all quality and speed claims — including the 50% latency reduction — are OpenAI's own figures and have not been independently verified.
The Images 2.5 release comes one week after OpenAI's GPT-6 Astra launch, marking a rapid succession of releases across the company's model lineup.
What this means
Images 2.5 is an incremental update rather than an architectural overhaul — OpenAI is iterating on editing precision and latency rather than introducing new capabilities like video or 3D generation. The bigger story for developers is the API split between Flare and Sunburst: OpenAI is explicitly segmenting its image API by use case (volume/speed vs. precision/quality) rather than shipping one general-purpose model, a pattern it has already applied to its text models. Without published pricing or third-party benchmarks, it's not yet possible to verify how the 50% latency claim holds up in production workloads, or how Sunburst's editing precision compares against competitors like Midjourney or Black Forest Labs' Flux models.
Related Articles
OpenAI Launches GPT-6 Astra With Half the Message Allowance of GPT-5.6 Sol
OpenAI has begun rolling out GPT-6 Astra to top-tier ChatGPT plans, the API, Azure, and AWS Bedrock. The model delivers roughly half the usage allowance of GPT-5.6 Sol across comparable plans, with Plus and Business users gaining access in the coming days.
OpenAI Lists GPT-6 Astra Pro on OpenRouter: Same Model, Higher-Compute Reasoning Mode
GPT-6 Astra Pro, now listed on OpenRouter, is the existing GPT-6 Astra model configured to run with reasoning.mode set to 'pro' for higher-quality output on complex tasks. It carries a 1M-token context window and tiered pricing from $5/$25 to $20/$100 per million input/output tokens depending on the serving tier.
OpenAI Claims Unreleased Model Solved Navier-Stokes Problem, Faces Credit Dispute
OpenAI claims an internal, unreleased model solved part of the Navier-Stokes Millennium Prize problem using roughly 10,000 concurrent agents over 88 hours. NYU mathematician Tristan Buckmaster has questioned whether the effort drew on his unpublished research with an Anthropic researcher.
Unverified 'GPT-6 Astra' Reportedly Completes Portal Solo in Under 24 Hours, No Official OpenAI Confirmation
A developer named cozyblaze posted on X that a model called 'GPT-6 Astra' completed Portal from start to finish without human intervention in 23 hours 43 minutes. OpenAI has not confirmed the existence of a model by that name, and all details come from a single third-party account.
Comments
Loading...