OpenAI's GPT Image 2 Outperforms Competitors in April 2026

OpenAI Strengthens Its Leadership Position

Since December, GPT Image 1.5 from OpenAI has already dominated the image generation sector. However, with the recent launch of GPT Image 2, the California-based firm has further solidified its lead. According to the latest ranking from Artificial Analysis, which relies on user votes to assess model performance, GPT Image 2 has taken the top spot with a record Elo score of 1,332. This model stands out with its Thinking mode, an innovation that allows it to search for up-to-date information on the web, thereby enhancing text rendering and photorealism. Artificial Analysis noted significant improvements in handling complex compositions, where other models often failed to follow instructions.

An example of an image generated with GPT Image 2 © OpenAI

The Leaders of the April 2026 Ranking

The April 2026 ranking, highly anticipated by industry observers, places GPT Image 2 at the top, followed by GPT Image 1.5 and Nano Banana 2 from Google, which had not managed to establish itself during its launch in February. ByteDance occupies the fifth position with Seedream 4.0, while Black Forest Labs stands out with its FLUX.2 models, which hold the sixth, eighth, and tenth places thanks to their FLUX.2 [max], FLUX.2 [pro], and FLUX.2 [flex] versions. A new player, Microsoft, makes a notable entry with MAI-Image 2, its second internally designed model, which ranks seventh by emphasizing photorealism. xAI, Elon Musk's company, positions itself in ninth place with its model grok-imagine-image.

Here are the 10 highest-performing image generation models as of April 2026:

• GPT Image 2 (high): 1,332 (Elo score)
• GPT Image 1.5 high (OpenAI): 1,271
• Nano Banana 2 (Google): 1,262
• Nano Banana Pro (Google): 1,217
• Seedream 4.0 (ByteDance): 1,204
• FLUX.2 [max] (Black Forest Labs): 1,203
• MAI-Image 2 (Microsoft): 1,196
• FLUX.2 [pro] (Black Forest Labs): 1,188
• grok-imagine-image (xAI): 1,186
• FLUX.2 [flex] (Black Forest Labs): 1,182

The Scoring System of Artificial Analysis

The ranking is based on the Text-to-image Arena of Artificial Analysis, a blind duel system where users compare two generated images without knowing the origin of the models. They choose the one that best meets the request, which allows for an Elo score to be assigned to each model. This score evolves over time, with a victory against a higher-ranked model increasing the score, while a defeat against a lower-ranked model decreases it. Artificial Analysis specifies that a high Elo score indicates a consistent user preference for a given model.