Anthropic Claude may have an Image generation model in the work

Anthropic, the AI safety-focused company behind the Claude family of large language models (LLMs), has made significant strides in multimodal AI. As of February 2026, the latest models are Claude Opus 4.5, Sonnet 4.5, and Haiku 4.5, offers advanced capabilities in text generation, reasoning, and coding. These models are multimodal, meaning they can process and analyze visual inputs such as photos, charts, diagrams, and documents alongside text.

Claude features allow users to extract insights from uploaded images, answer questions about visual content, and support tasks like document analysis or interpreting graphs. Users can upload images for deeper contextual understanding, and the models can generate visual study aids through artifacts, such as diagrams or charts produced via code execution.

However, Anthropic’s current Claude models do not include native image generation capabilities. Official documentation and announcements consistently state that Claude outputs are limited to text (with indirect visual elements possible through code-generated artifacts, such as SVGs or plots).

Unlike competitors such as OpenAI’s DALL-E or Grok Imagine, Claude cannot directly create images from text prompts. Anthropic has emphasized this text-only output restriction in discussions of safety, noting it reduces risks associated with synthetic media.

There has been no official announcement from Anthropic regarding a dedicated text-to-image model or native image creation feature in upcoming releases. The company continues to prioritize alignment, safety, and incremental improvements in reasoning and multimodal understanding.

A leaked configuration file from Anthropic’s Claude AI reveals built-in features for image creation and editing. The code includes functions named “create_image” and “edit_image,” along with simple user messages like “Created an image,” “Created {count} images,” “Creating image,” and “Editing image.”

This snippet confirms that Claude will soon support direct image generation and modification within the app. A related model called Sonata has appeared on the LMSYS Chatbot Arena, where it inconsistently claims to be developed by either Anthropic or Google.

Anthropic may be taking help from Google to offer an image generation model. However, it may not be native, at least from the beginning. More information about this is yet to come.

Source – @btibor91/X

(source)