Claude Opus 4.7 Takes the Top Spots in Slide Generation
Anthropic's "Claude Opus 4.7" and "Claude Opus 4.7 (Thinking)" have taken the No. 1 and No. 2 spots in the Slides category of "Design Arena," a crowdsourced benchmark that evaluates AI-generated design, pulling ahead of other models by more than 80 Elo, as revealed in an official X post by the operator (June 1, 2026). 1
What Happened
According to a post by the Design Arena official account (@Designarena), in the Slides category Anthropic's two models lead other models by more than 80 points in Elo rating, establishing a de facto standard for slide generation. Following from third place are Z.ai's "GLM 5.1," Google DeepMind's "Gemini 3.5 Flash," and Moonshot's "Kimi K2.6." 1
Design Arena is a platform where users vote on side-by-side generation results from multiple models for the same prompt, updating rankings via an Elo system. The Slides category targets "Agentic Slides" (autonomous slide generation and editing) for evaluation, focusing on PowerPoint-style PPTX output and presentation material generation. 2
Background and Significance
Design Arena bills itself as "the world's first crowdsourced benchmark for AI-generated design," operated by The Intelligence Company / Arcada Labs, and has multiple categories including UI/Website, Image, 3D, Code, and Slides. The Slides Arena was revived around March 2026, and dedicated tools such as Gamma, Alai, and Manus also participate. 3
Anthropic added "Claude Opus 4.7" to Design Arena around April 16, 2026. The following day, on the 17th, it released a new product based on the Opus 4.7 foundation model, "Claude Design" (Anthropic Labs), as a research preview. It is a tool that can conversationally generate and edit prototypes, slides, and one-pagers from text prompts, screenshots, and existing PPTX/DOCX files, and Anthropic positions the model as "best for visual work (design, prototypes, slides)." 4
Opus 4.7 has been rated as having enhanced visual reasoning, layout consistency, and long-horizon agentic workflows, and the technical blog Towards AI noted that "Claude Design is the bigger story for Opus 4.7." Specific improvements such as 69%→82% on a visual reasoning benchmark have also been reported. Anthropic models continue to show a strong tendency in "soft-verifiable, creative domains" where a clear correct answer is hard to verify mechanically. 5
Reactions
While engagement on the originating post itself is on the lower side, past posts of the same kind (such as on May 15) drew a certain amount of attention with over 250 Likes, and the configuration of Opus 4.7 at No. 1 and the Thinking version at No. 2 has been continuously confirmed. 6
As a positive use case, posts on X touting "Goodbye, PowerPoint" and sharing workflows and prompt collections for creating professional-grade slides within 60 seconds with Claude 4.7 are active. Examples are also introduced of uploading existing materials to automatically generate 19 brand-matched slides and self-correcting the layout. 7
On the other hand, there are also remarks about cost and speed, such as "Claude is expensive" and "Kimi is slow." GLM 5.1 users have voiced that "Claude is expensive, but GLM is cheap and a superior alternative at the top of the Slides Arena," and the trade-off between price and performance is being debated. Also, in the 3D design category, there are reports of open-source models such as "Kimi K2.6" occupying top spots, making the difference between slide specialization and general-purpose versatility a point of discussion. 8