Moonshot AI shows Kimi Slides generating editable PowerPoint decks

Moonshot AI demonstrated Kimi Slides, a workflow that turns research and source material into an editable PowerPoint presentation, in an August 3 tutorial on X. The feature, powered by the 2.8-trillion-parameter Kimi K3 model, handles outline, research, visual design, and citations, and is available inside Kimi Work. Moonshot AI CEO Yang Zhilin, who co-authored Transformer-XL and XLNet, positions the tool as a concrete output for the company's long-context AI research.

Yang Zhilin https://kimiyoung.github.io/?ref=runtimewire 's Moonshot AI https://www.moonshot.ai/?ref=runtimewire used an August 3 tutorial on X https://x.com/Kimi Moonshot/status/2084245860339298423?ref=runtimewire to demonstrate Kimi Slides https://www.kimi.com/en/features/slides?ref=runtimewire , a workflow that turns research and source material into a designed, editable presentation inside Kimi Work https://www.kimi.com/products/kimi-work?ref=runtimewire . Kimi on X https://x.com/Kimi Moonshot/status/2084245860339298423?ref=runtimewire The tutorial shows Kimi handling the outline, research and visual design, including charts and SmartArt, before producing a deck that can be downloaded and edited. Moonshot AI is promoting an existing feature rather than announcing its launch: the tutorial establishes what Kimi Slides can do today, while the feature's original release date remains unclear. For Yang, the presentation generator puts a familiar office deliverable around the technical ideas he has pursued for years. He earned his bachelor's degree from Tsinghua University in 2015 and a computer science PhD from Carnegie Mellon University in 2019, according to his personal site https://kimiyoung.github.io/?ref=runtimewire . He also worked at Google Brain and Meta AI and co-authored Transformer-XL and XLNet, two influential approaches to language modeling and long-context processing. That background helps explain Moonshot AI's product path. Yang has built Kimi around models that can process large bodies of material, reason across them and produce finished work. A slide deck gives that research agenda a concrete destination: a file that a consultant, analyst, founder or teacher can revise, present and send to someone else. From a prompt to a PowerPoint file Kimi's materials describe Kimi Slides https://www.kimi.com/en/features/slides?ref=runtimewire as a presentation generator that can start from a topic, pasted text or uploaded files, then research, structure, design and export an editable PowerPoint deck. Supported uploads include images, PDFs, Word documents, spreadsheets and existing PowerPoint templates. Moonshot AI also says Kimi can generate on-slide citations. Moonshot AI's pitch emphasizes editable output rather than static screenshots: decks can be downloaded as PowerPoint files and revised. The feature also supports PowerPoint templates, which matters for teams that need brand-consistent output. For a researched deck with citations and charts, the useful workflow is an assigned task the user checks before sharing. It changes the workflow from real-time assistance to a draft that the user reviews later. Moonshot AI's claims about citations put a clear burden on the output. The product materials https://www.kimi.com/help/slides/ppt-overview?ref=runtimewire say Kimi can generate source citations. Users still have to inspect those citations, the interpretation of the underlying material and any numbers that move from a source document into a chart. An editable PowerPoint file makes correction easier; it does not make the model's research automatically reliable. Kimi K3 becomes a product layer The slide workflow is powered by Kimi K3 https://www.kimi.com/blog/kimi-k3?ref=runtimewire , which Moonshot AI introduced in July. Moonshot AI describes K3 as a 2.8-trillion-parameter, native multimodal model with a 1-million-token context window. Moonshot AI has published open weights and technical materials on GitHub https://github.com/MoonshotAI/Kimi-K3?ref=runtimewire . Those model claims are especially relevant in presentations, where an agent has to decide which facts deserve emphasis, how a chart should frame them and which conclusions belong in the final narrative. Kimi Slides sits inside Kimi Work https://www.kimi.com/products/kimi-work?ref=runtimewire , Moonshot AI's desktop agent for knowledge workers. Kimi Work can connect with local files, automate browser tasks through WebBridge, schedule recurring work and turn research into presentations or spreadsheets. Slides therefore serve as one output within a broader attempt to automate an office assignment from collection through delivery. That integration is Moonshot AI's strongest product argument. A standalone presentation generator begins with material the user has already gathered. Kimi Work is designed to collect material from files and the web, analyze it and pass the result into the deck generator. The quality of the final presentation will depend on whether that chain preserves sources, context and user intent across each step. A funded lab enters a proven category Moonshot AI has ample reason to package K3 into paid workflows. TechCrunch reported on May 7 https://techcrunch.com/2026/05/07/chinas-moonshot-ai-raises-2b-at-20b-valuation-as-demand-for-open-source-ai-skyrockets/?ref=runtimewire that Moonshot AI raised about $2 billion at a $20 billion valuation in a round led by Meituan's investment arm, Long-Z Investments. Tsinghua Capital, China Mobile and CPE Yuanfeng also participated, according to the report. The same report said Moonshot AI's annual recurring revenue topped $200 million in April, citing a post from a financial adviser to participating investors. Paid subscriptions and API use drove that figure. Kimi Slides offers another way to convert model usage into recurring work inside Moonshot AI's own interface, where Moonshot AI controls the customer relationship and the surrounding tools. Presentation generation is already a competitive software category. Gamma said in November 2025 https://gamma.app/insights/how-we-built-a-usd100m-business-differently?ref=runtimewire that it had crossed $100 million in annual recurring revenue, reached 70 million users and raised $68 million at a $2.1 billion valuation. Those figures are self-reported. Beautiful.ai said in March 2026 https://www.beautiful.ai/blog/beautiful-ai-launches-context-aware-ai-workflow-and-secures-45m-investment-from-general-catalyst-to-accelerate-global-growth?ref=runtimewire that it secured a $45 million investment from General Catalyst while introducing a context-aware presentation workflow. Moonshot AI is entering that contest with a different starting point. Gamma and Beautiful.ai built focused presentation products and expanded their AI capabilities. Yang built a foundation-model lab and is now pushing K3 into office software. Kimi Slides will test whether Moonshot AI's model research can produce a deck that users trust, edit and reuse, rather than merely a polished demonstration. The August 3 tutorial is a modest piece of product education. It also shows the commercial shape of Yang's bet: Kimi K3 earns its place on a knowledge worker's desktop by finishing recognizable work. For presentation software, that means sound research, defensible citations and a file that survives the user's edits after the model is done.