OpenAI announced the general availability of its latest image generation model through the API on 23 April 2025 [source]. The model, designated DALL·E 3, is now accessible to developers via the Images API endpoint with pricing set at $0.040 per image for standard 1024×1024 resolution.

The provider states the model offers "improved prompt adherence" and "higher fidelity" compared to the previous DALL·E 2 release. OpenAI claims the system better interprets complex prompts with multiple objects, spatial relationships, and specific artistic styles. The model supports three resolution options: 1024×1024, 1024×1792, and 1792×1024 pixels.

According to the announcement, DALL·E 3 incorporates safety mitigations including content filtering for policy violations and a system to decline requests that ask for images in the style of living artists. The provider reports the model was trained with an emphasis on reducing bias in generated outputs, though specific training data composition remains undisclosed.

Rate limits are set at 5 requests per minute for standard tier users, with higher limits available for enterprise customers. The API returns images as base64-encoded JSON or direct URLs with 60-minute expiration windows.

OpenAI has deprecated the DALL·E 2 endpoint effective immediately, with existing integrations required to migrate within 90 days. The provider notes that DALL·E 3 was previously available only through ChatGPT Plus and Enterprise subscriptions since October 2024.

No independent benchmarks were cited in the announcement. The model's performance characteristics, failure modes, and drift behaviour under production load remain to be observed by third-party users.