Recraft
High-resolution, design-focused text-to-image generation model from Recraft, available through Recraft Studio and the Recraft API with raster and vector variants.
Recraft V4.1 Pro is a specialized image model rather than a general-purpose language or agent model. Official Recraft materials support high-resolution, design-focused text-to-image generation through Studio and API, with raster and vector variants. Artificial Analysis reports an Image Arena Intelligence Elo of 1155.11, rank 14 of 81, which supports a solid technical score within image generation. That is materially below broad leaders such as Claude Opus 4.8 and GPT-5.5 on general technical breadth, coding, and reasoning, but stronger than a low-end general-model anchor for its focused visual-production use case. Coding and agentic-work scores are intentionally low: DeepSWE and LiveCodeBench list no result, SWE-bench evidence was not found, and supplied material describes an image generator rather than a coding agent. Reasoning is likewise not established by the available evidence. Multimodal I/O is comparatively strong because the product exposes image generation in both raster and vector-oriented workflows, though no evidence establishes broad input modalities or tool-use depth comparable with Gemini 3.5 Flash. Developer experience benefits from documented API access and a Studio surface. Cost effectiveness is only provisional: Artificial Analysis tracks per-1,000-image pricing and generation time, but the supplied evidence gives neither the actual price nor latency figure. Pricing clarity is therefore below published API-model anchors. Public signal is moderate, supported by Artificial Analysis coverage and Text-to-Image Arena inclusion, not a reported arena rank. Evidence quality is reasonable for product existence and positioning, but benchmark coverage is narrow and mostly modality-specific.
