Get API access
Model release

GPT Image 2.5 Sunburst and Flare Release Details

OpenAI released GPT Image 2.5 Sunburst and Flare on September 8, 2026. This brief covers the documented API endpoints, new quality settings, and evaluation…

Released Published 3 min readModel evidence ↗
GPT Image 2.5 Sunburst release illustration — OpenAI

Release Announcement and Model Scope

The official changelog entry is dated September 8, 2026. This date marks the historical event of the release announcement. The release includes two specific model identifiers: gpt-image-2.5-sunburst and gpt-image-2.5-flare. Both models are categorized under the GPT family. They accept text and image inputs and produce image outputs. The source does not provide context window sizes or maximum output token limits for these specific versions.

  • GPT Image 2.5 Sunburst is designated for workflows where editing precision matters most.
  • GPT Image 2.5 Flare is intended for fast, high-quality everyday image generation.
  • Both models operate through the v1/images API and the v1/responses API image generation tool.

Quality Settings and Pricing Information

The documentation states that both new models support two additional quality settings: xhigh and max. These are described as new options for these specific releases. Regarding cost, the source confirms that both models use GPT Image 2 token rates. No separate pricing structure or unique rate card was established for these specific versions in the provided facts. Specific numerical pricing values are not detailed in the release summary.

  • New quality levels include xhigh and max.
  • Token rates align with the existing GPT Image 2 standard.
  • Specific numerical pricing values are not detailed in the release summary.

Evaluation Procedure and Evidence Limits

To evaluate these models, engineers should design tests based on the stated use cases. For Sunburst, create a test suite involving complex image editing tasks to measure precision against previous baselines. For Flare, generate a batch of standard images to measure speed and visual quality. Compare the results from the xhigh and max settings against lower-tier defaults to validate the impact of the new quality parameters. The source records that open weights are not confirmed. It does not establish whether local installation is possible or if API access is mandatory. Evaluation must occur via the API endpoints listed in the changelog.

  • Test Sunburst with multi-step editing prompts to assess precision.
  • Measure generation latency for Flare using standard daily-use prompts.
  • Compare output fidelity between xhigh/max settings and standard settings.
  • Validate API response times using the v1/images and v1/responses endpoints.

Primary sources

QUICK ANSWERS

Frequently asked questions

What is the difference between Sunburst and Flare?

Sunburst is optimized for workflows requiring high editing precision, whereas Flare is designed for fast, high-quality everyday image generation. The choice depends on whether the primary need is detailed modification or rapid creation.

Are the weights for these models open?

The source indicates that open weights are not confirmed for this release. The documentation does not specify if local installation is possible or if API access is mandatory.