Claude Opus 4.7 Dominates Design Arena's Slides Category
It has come to light that Anthropic's latest flagship "Claude Opus 4.7" and its extended-thinking version "Opus 4.7 (Thinking)" lead the Slides (slide generation) category of the AI design evaluation platform Design Arena by a wide margin, holding a lead of more than 80 Elo points over the second-place and lower models.1
What Happened
On Design Arena's Slides leaderboard, Claude Opus 4.7 and Claude Opus 4.7 (Thinking) (@AnthropicAI) occupy the top spots, holding a lead of more than 80 Elo points over trailing models such as GLM 5.1, Gemini 3.5 Flash, and Kimi K2.6.1
Design Arena is a crowdsourced benchmark that compares each model's outputs to the same prompt via user voting and computes ratings using the Elo method. In the Slides category, the quality of slide generation is the subject of evaluation.7
The underlying Claude Opus 4.7 is the flagship model for which Anthropic began general availability (GA) on April 16, 2026, with the API model ID claude-opus-4-7. Pricing is $5 input / $25 output per 1 million tokens, and the context window is said to be 1M tokens.2
Background and Significance
Anthropic positions Opus 4.7 as able to generate "higher-quality interfaces, slides, and documents." What is driving up slide generation quality in particular is high-resolution vision: the long edge of image inputs reaches up to 2,576 pixels (about 3.75 megapixels, more than triple the previous figure), improving the accuracy of recognizing fine details in text and diagrams.2
Furthermore, shortly after release, Anthropic introduced Claude Design (research preview), a tool that generates visual deliverables from prompts based on Opus 4.7. Product expansion of slide and prototype generation is also advancing, including brand application and enhanced output to PDF / PPTX / Canva.3
In the slides domain, dedicated tools such as Gamma, Alai, and Manus, as well as other LLMs, compete, but Anthropic's models clearly lead in Design Arena's Slides category. They showed strength even at the previous-generation Opus 4.6, and with 4.7 the gap has widened further.4
Reactions
On X, evaluations such as "Claude is expensive but the strongest for Slides" stand out, while there are also practical voices saying, "It's pricey, so depending on the use case I use it alongside GLM 5.1 and the like." There are also observations that Chinese open-weight models such as Kimi and GLM are doing well in other categories such as 3D.5
Concrete use cases cited include converting PDFs into brand-applied slides, generating large-scale decks from research papers, and creating prototypes. On the other hand, stumbling points have been reported, such as "the benefits of high-resolution vision are large, but prompt tuning may be needed in some cases."2
In Anthropic's official announcement, early testers such as Hex, Notion, and Harvey reported significant improvements in document- and slide-related tasks.2