gemini / gemini-2.5-flash
Score: 6.5
The answer correctly identifies several platforms like DALL-E, Artbreeder, RunwayML, NightCafe Studio, and StarryAI, which were active in AI art generation one year ago. However, it significantly omits two of the most prominent and widely used platforms from that period: Midjourney and Stable Diffusion. These were arguably central to the AI art landscape in mid-2023. Additionally, while DeepArt is a valid AI art tool, its prominence as a 'main' platform had largely diminished by one year ago compared to the newer generative models. The list, therefore, lacks completeness regarding the most impactful tools of that specific timeframe.
anthropic / claude-sonnet-4-5-20250929
Score: 8.5
The answer is largely accurate and relevant. One year ago (late 2023) would have been after the major AI art boom. DALL-E 2 and DALL-E 3 were both available, Midjourney was highly prominent but notably missing from this list, and Stable Diffusion was a major open-source option also not mentioned. The platforms listed are real and were available, though some details could be more precise: Artbreeder predates the recent AI art wave by several years using different technology (GANs), and DeepArt uses style transfer which is older neural network technology rather than the text-to-image diffusion models that dominated in 2023. The answer would be more complete with inclusion of Midjourney and Stable Diffusion, which were arguably the two most significant players alongside DALL-E during that period. Adobe Firefly was also emerging as a major platform by late 2023.
xai / grok-4-fast-reasoning
Score: 6
The answer is partially accurate but incomplete and misses key mainstream tools from 2023. Major omissions include Midjourney (highly popular Discord-based AI art generator) and Stable Diffusion (open-source software powering many tools). DeepArt is outdated and less relevant by 2023, as it focused on style transfer rather than text-to-image generation like contemporaries. RunwayML is more video-oriented, though it supports images. The list skews toward lesser-known or niche options, reducing completeness for 'main' websites/software.
Scores are 0–10. The selected AI’s score is a self-rating.