Abacus AI Studio demos 100 models

Creative AI is getting bundled. That’s the real story here. Abacus AI’s new Abacus Studio is not one new image model or one new video model — it’s a front end for a whole pile of them, plus some agent-style automation that tries to hide the model-picking work from the user. That matters because the annoying part of generative media now isn’t just writing prompts. It’s knowing which model to use, when to switch, and how to iterate fast enough to get something usable. Abacus is trying to turn that mess into one workflow. ### What actually launched? Abacus Studio is a chat-style workspace for generating, editing, and enhancing images and videos, with speech tools folded in too. The company pitches it as one place to create visuals instead of bouncing between separate apps for image generation, video generation, lip sync, upscaling, and editing. On the public product pages, Abacus describes Studio as “all-in-one” and shows doing. ### Why are people saying “100 models”? That number is a little slippery. Abacus’s broader ChatLLM product says it gives users access to “100+ top language, video and image models.” But the Studio pages themselves use smaller counts in public-facing copy — “50+” on one page and “40+” on another — while the FAQ lists a long but still finite lineup of image, video, and speech models. So the clean read is this overall, while Abacus Studio is the media-focused slice of that stack. ### What models are actually inside it? The lineup is the part with real weight. Abacus lists image tools like GPT Image 2, Midjourney, Imagen 4, FLUX.2 Pro, Ideogram 3.0, Recraft SVG, and Magnific. On video, it lists Sora 2, Veo 3.1, Seedance 2.0, Kling AI v3, Hailuo 2, Luma Labs, and Topaz Upscaler. Speech support includes ElevenLabs, OpenAI, and Hume. Basically, Abacus is trying to be the control panel above the models, not the one model you commit to. ### What does the “agentic” part really mean? It mostly means automation and routing, not magic. Abacus says Auto Mode chooses the best model, resolution, and settings based on the prompt — static prompts go toward image generation, motion-heavy prompts go toward video. Elsewhere in Abacus’s docs, the company describes its AI agent as something that can complete end-to-end tasks and package outputs for use, describe the result, let the system decide the path.” ### Why is that useful? Because model choice has become the new prompt engineering. One model does faces better. Another handles motion better. Another is better at edits than first-pass generation. A creative workflow now looks less like asking one model for a masterpiece and more like moving a project through stages — generate, edit, upscale, lip-sync, revise. Abacus is packaging that as a single workspace subscriptions and ten exports. ### What’s the catch? The catch